Journal tags: performance



Doing the right thing for the wrong reasons

I remember trying to convince people to use semantic markup because it’s good for accessibility. That tactic didn’t always work. When it didn’t, I would add “By the way, Google’s searchbot is indistinguishable from a screen-reader user so semantic markup is good for SEO.”

That usually worked. It always felt unsatisfying though. I don’t know why. It doesn’t matter if people do the right thing for the wrong reasons. The end result is what matters. But still. It never felt great.

It happened with responsive design and progressive enhancement too. If I couldn’t convince people based on user experience benefits, I’d pull up some official pronouncement from Google recommending those techniques.

Even AMP, a dangerously ill-conceived project, has one very handy ace in the hole. You can’t add third-party JavaScript cruft to AMP pages. That’s useful:

Beleaguered developers working for publishers of big bloated web pages have a hard time arguing with their boss when they’re told to add another crappy JavaScript tracking script or bloated library to their pages. But when they’re making AMP pages, they can easily refuse, pointing out that the AMP rules don’t allow it. Google plays the bad cop for us, and it’s a very valuable role.

AMP is currently dying, which is good news. Google have announced that core web vitals will be used to boost ranking instead of requiring you to publish in their proprietary AMP format. The really good news is that the political advantage that came with AMP has also been ported over to core web vitals.

Take user-hostile obtrusive overlays. Perhaps, as a contientious developer, you’ve been arguing for years that they should be removed from the site you work on because they’re so bad for the user experience. Perhaps you have been met with the same indifference that I used to get regarding semantic markup.

Well, now you can point out how those annoying overlays are affecting, for example, the cumulative layout shift for the site. And that number is directly related to SEO. It’s one thing for a department to over-ride UX concerns, but I bet they’d think twice about jeopardising the site’s ranking with Google.

I know it doesn’t feel great. It’s like dealing with a bully by getting an even bigger bully to threaten them. Still. Needs must.


Core web vitals from Google are the ingredients for an alphabet soup of exlusionary intialisms. But once you get past the unnecessary jargon, there’s a sensible approach underpinning the measurements.

From May—no, June—these measurements will be a ranking signal for Google search so performance will become more of an SEO issue. This is good news. This is what Google should’ve done years ago instead of pissing up the wall with their dreadful and damaging AMP project that blackmailed publishers into using a proprietary format in exchange for preferential search treatment. It was all done supposedly in the name of performance, but in reality all it did was antagonise users and publishers alike.

Core web vitals are an attempt to put numbers on user experience. This is always a tricky balancing act. You’ve got to watch out for the McNamara fallacy. Harry has already started noticing this:

A new and unusual phenomenon: clients reluctant (even refusing) to fix performance issues unless they directly improve Vitals.

Once you put a measurement on something, there’s a danger of focusing too much on the measurement. Chris is worried that we’re going to see tips’n’tricks for gaming core web vitals:

This feels like the start of a weird new era of web performance where the metrics of web performance have shifted to user-centric measurements, but people are implementing tricky strategies to game those numbers with methods that, if anything, slightly harm user experience.

The map is not the territory. The numbers are a proxy for user experience, but it’s notoriously difficult to measure intangible ideas like pain and frustration. As Laurie says:

This is 100% the downside of automatic tools that give you a “score”. It’s like gameification. It’s about hitting that perfect score instead of the holistic experience.

And Ethan has written about the power imbalance that exists when Google holds all the cards, whether it’s AMP or core web vitals:

Google used its dominant position in the marketplace to force widespread adoption of a largely proprietary technology for creating websites. By switching to Core Web Vitals, those power dynamics haven’t materially changed.

We would do well to remember:

When you measure, include the measurer.

But if we’re going to put numbers to user experience, the core web vitals are a pretty good spread of measurements: largest contentful paint, cumulative layout shift, and first input delay.

(If you prefer using initialisms, remember that CFP is Certified Financial Planner, CLS is Community Legal Services, and FID is Flame Ionization Detector. Together they form CWV, Catholic War Veterans.)

Cascading Style Sheets

There are three ways—that I know of—to associate styles with markup.

External CSS

This is probably the most common. Using a link element with a rel value of “stylesheet”, you point to a URL using the href attribute. That URL is a style sheet that is applied to the current document (“the relationship of the linked resource it is that is a ‘stylesheet’ for the current document”).

<link rel="stylesheet" href="/path/to/styles.css">

In theory you could associate a style sheet with a document using an HTTP header, but I don’t think many browsers support this in practice.

You can also pull in external style sheets using the @import declaration in CSS itself, as long as the @import rule is declared at the start, before any other styles.

@import url('/path/to/more-styles.css');

When you use link rel="stylesheet" to apply styles, it’s a blocking request: the browser will fetch the style sheet before rendering the HTML. It needs to know how the HTML elements will be painted to the screen so there’s no point rendering the HTML until the CSS is parsed.

Embedded CSS

You can also place CSS rules inside a style element directly in the document. This is usually in the head of the document.

element {
    property: value;

When you embed CSS in the head of a document like this, there is no network request like there would be with external style sheets so there’s no render-blocking behaviour.

You can put any CSS inside the style element, which means that you could use embedded CSS to load external CSS using an @import statement (as long as that @import statement appears right at the start).

@import url('/path/to/more-styles.css');
element {
    property: value;

But then you’re back to having a network request.

Inline CSS

Using the style attribute you can apply CSS rules directly to an element. This is a universal attribute. It can be used on any HTML element. That doesn’t necessarily mean that the styles will work, but your markup is never invalidated by the presence of the style attribute.

<element style="property: value">

Whereas external CSS and embedded CSS don’t have any effect on specificity, inline styles are positively radioactive with specificity. Any styles applied this way are almost certain to over-ride any external or embedded styles.

You can also apply styles using JavaScript and the Document Object Model. = 'value';

Using the DOM style object this way is equivalent to inline styles. The radioactive specificity applies here too.

Style declarations specified in external style sheets or embedded CSS follow the rules of the cascade. Values can be over-ridden depending on the order they appear in. Combined with the separate-but-related rules for specificity, this can be very powerful. But if you don’t understand how the cascade and specificity work then the results can be unexpected, leading to frustration. In that situation, inline styles look very appealing—there’s no cascade and everything has equal specificity. But using inline styles means foregoing a lot of power—you’d be ditching the C in CSS.

A common technique for web performance is to favour embedded CSS over external CSS in order to avoid the extra network request (at least for the first visit—there are clever techniques for caching an external style sheet once the HTML has already loaded). This is commonly referred to as inlining your CSS. But really it should be called embedding your CSS.

This language mix-up is not a hill I’m going to die on (that hill would be referring to blog posts as blogs) but I thought it was worth pointing out.

Clean advertising

Imagine if you were told that fossil fuels were the only way of extracting energy. It would be an absurd claim. Not only are other energy sources available—solar, wind, geothermal, nuclear—fossil fuels aren’t even the most effecient source of energy. To say that you can’t have energy without burning fossil fuels would be pitifully incorrect.

And yet when it comes to online advertising, we seem to have meekly accepted that you can’t have effective advertising without invasive tracking. But nothing could be further from the truth. Invasive tracking is to online advertising as fossil fuels are to energy production—an outmoded inefficient means of getting substandard results.

Before the onslaught of third party cookies and scripts, online advertising was contextual. If I searched for property insurance, I was likely to see an advertisement for property insurance. If I was reading an article about pet food, I was likely to be served an advertisement for pet food.

Simply put, contextual advertising ensured that the advertising that accompanied content could be relevant and timely. There was no big mystery about it: advertisers just needed to know what the content was about and they could serve up the appropriate advertisement. Nice and straightforward.

Too straightforward.

What if, instead of matching the advertisement to the content, we could match the advertisement to the person? Regardless of what they were searching for or reading, they’d be served advertisements that were relevant to them not just in that moment, but relevant to their lifestyles, thoughts and beliefs? Of course that would require building up dossiers of information about each person so that their profiles could be targeted and constantly updated. That’s where cross-site tracking comes in, with third-party cookies and scripts.

This is behavioural advertising. It has all but elimated contextual advertising. It has become so pervasive that online advertising and behavioural advertising have become synonymous. Contextual advertising is seen as laughably primitive compared with the clairvoyant powers of behavioural advertising.

But there’s a problem with behavioural advertising. A big problem.

It doesn’t work.

First of all, it relies on mind-reading powers by the advertising brokers—Facebook, Google, and the other middlemen of ad tech. For all the apocryphal folk tales of spooky second-guessing in online advertising, it mostly remains rubbish.

Forget privacy: you’re terrible at targeting anyway:

None of this works. They are still trying to sell me car insurance for my subway ride.

Have you actually paid attention to what advertisements you’re served? Maciej did:

I saw a lot of ads for GEICO, a brand of car insurance that I already own.

I saw multiple ads for Red Lobster, a seafood restaurant chain in America. Red Lobster doesn’t have any branches in San Francisco, where I live.

Finally, I saw a ton of ads for Zipcar, which is a car sharing service. These really pissed me off, not because I have a problem with Zipcar, but because they showed me the algorithm wasn’t even trying. It’s one thing to get the targeting wrong, but the ad engine can’t even decide if I have a car or not! You just showed me five ads for car insurance.

And yet in the twisted logic of ad tech, all of this would be seen as evidence that they need to gather even more data with even more invasive tracking and surveillance.

It turns out that bizarre logic is at the very heart of behavioural advertising. I highly recommend reading the in-depth report from The Correspondent called The new dot com bubble is here: it’s called online advertising:

It’s about a market of a quarter of a trillion dollars governed by irrationality.

The benchmarks that advertising companies use – intended to measure the number of clicks, sales and downloads that occur after an ad is viewed – are fundamentally misleading. None of these benchmarks distinguish between the selection effect (clicks, purchases and downloads that are happening anyway) and the advertising effect (clicks, purchases and downloads that would not have happened without ads).

Suppose someone told you that they keep tigers out of their garden by turning on their kitchen light every evening. You might think their logic is flawed, but they’ve been turning on the kitchen light every evening for years and there hasn’t been a single tiger in the garden the whole time. That’s the logic used by ad tech companies to justify trackers.

Tracker-driven behavioural advertising is bad for users. The advertisements are irrelevant most of the time, and on the few occasions where the advertising hits the mark, it just feels creepy.

Tracker-driven behavioural advertising is bad for advertisers. They spend their hard-earned money on invasive ad tech that results in no more sales or brand recognition than if they had relied on good ol’ contextual advertising.

Tracker-driven behavioural advertising is very bad for the web. Megabytes of third-party JavaScript are injected at exactly the wrong moment to make for the worst possible performance. And if that doesn’t ruin the user experience enough, there are still invasive overlays and consent forms to click through (which, ironically, gets people mad at the legislation—like GDPR—instead of the underlying reason for these annoying overlays: unnecessary surveillance and tracking by the site you’re visiting).

Tracker-driven behavioural advertising is good for the middlemen doing the tracking. Facebook and Google are two of the biggest players here. But that doesn’t mean that their business models need to be permanently anchored to surveillance. The very monopolies that make them kings of behavioural advertising—the biggest social network and the biggest search engine—would also make them titans of contextual advertising. They could pivot from an invasive behavioural model of advertising to a privacy-respecting contextual advertising model.

The incumbents will almost certainly resist changing something so fundamental. It would be like expecting an energy company to change their focus from fossil fuels to renewables. It won’t happen quickly. But I think that it may eventually happen …if we demand it.

In the meantime, we can all play our part. Just as we can do our bit for the environment at an individual level by sorting our recycling and making green choices in our day to day lives, we can all do our bit for the web too.

The least we can do is block third-party cookies. Some browsers are now doing this by default. That’s good.

Blocking third-party JavaScript is a bit trickier. That requires a browser extension. Most of these extensions to block third-party tracking are called ad blockers. That’s a shame. The issue is not with advertising. The issue is with tracking.

Alas, because this software is labelled under ad blocking, it has led to the ludicrous situation of an ethical argument being made to allow surveillance and tracking! It goes like this: websites need advertising to survive; if you block the ads, then you are denying these sites revenue. That argument would make sense if we were talking about contextual advertising. But it makes no sense when it comes to behavioural advertising …unless you genuinely believe that online advertising has to be behavioural, which means that online advertising has to track you to be effective. Such a belief would be completely wrong. But that doesn’t stop it being widely held.

To argue that there is a moral argument against blocking trackers is ridiculous. If anything, there’s a moral argument to be made for installing anti-tracking software for yourself, your friends, and your family. Otherwise we are collectively giving up our privacy for a business model that doesn’t even work.

It’s a shame that advertisers will lose out if tracking-blocking software prevents their ads from loading. But that’s only going to happen in the case of behavioural advertising. Contextual advertising won’t be blocked. Contextual advertising is also more lightweight than behavioural advertising. Contextual advertising is far less creepy than behavioural advertising. And crucially, contextual advertising works.

That shouldn’t be a controversial claim: the idea that people would be interested in adverts that are related to the content they’re currently looking at. The greatest trick the ad tech industry has pulled is convincing the world that contextual relevance is somehow less effective than some secret algorithm fed with all our data that’s supposed to be able to practically read our minds and know us better than we know ourselves.

Y’know, if this mind-control ray really could give me timely relevant adverts, I might possibly consider paying the price with my privacy. But as it is, YouTube still hasn’t figured out that I’m not interested in Top Gear or football.

The next time someone is talking about the necessity of advertising on the web as a business model, ask for details. Do they mean contextual or behavioural advertising? They’ll probably laugh at you and say that behavioural advertising is the only thing that works. They’ll be wrong.

I know it’s hard to imagine a future without tracker-driven behavioural advertising. But there are no good business reasons for it to continue. It was once hard to imagine a future without oil or coal. But through collective action, legislation, and smart business decisions, we can make a cleaner future.

Portals and giant carousels

I posted something recently that I think might be categorised as a “shitpost”:

Most single page apps are just giant carousels.

Extreme, yes, but perhaps there’s a nugget of truth to it. And it seemed to resonate:

I’ve never actually seen anybody justify SPA transitions with actual business data. They generally don’t seem to increase sales, conversion, or retention.

For some reason, for SPAs, managers are all of a sudden allowed to make purely emotional arguments: “it feels snappier”

If businesses were run rationally, when somebody asks for an order of magnitude increase in project complexity, the onus would be on them to prove that it proportionally improves business results.

But I’ve never actually seen that happen in a software business.

A single page app architecture makes a lot of sense for interaction-heavy sites with lots of state to maintain, like But I’ve seen plenty of sites built as single page apps even though there’s little to no interactivity or state management. For some people, it’s the default way of building anything on the web, even a brochureware site.

It seems like there’s a consensus that single page apps may have long initial loading times, but then they have quick transitions between “pages” …just like a carousel really. But I don’t know if that consensus is based on reality. Whether you’re loading a page of HTML or loading a chunk of JSON, you’re still making a network request that will take time to resolve.

The argument for loading a chunk of JSON is that you don’t have to make any requests for the associated CSS and JavaScript—they’re already loaded. Whereas if you request a page of HTML, that HTML will also request CSS and JavaScript.

Leaving aside the fact that is literally what the browser cache takes of, I’ve seen some circular reasoning around this:

  1. We need to create a single page app because our assets, like our JavaScript dependencies, are so large.
  2. Why are the JavaScript dependencies so large?
  3. We need all that JavaScript to create the single page app functionlity.

To be fair, in the past, the experience of going from page to page used to feel a little herky-jerky, even if the response times were quick. You’d get a flash of a white blank page between navigations. But that’s no longer the case. Browsers now perform something called “paint holding” which elimates the herky-jerkiness.

So now if your pages are a reasonable size, there’s no practical difference in user experience between full page refreshes and single page app updates. Navigate around The Session if you want to see paint holding in action. Switching to a single page app architecture wouldn’t improve the user experience one jot.


If I were controlling everything with JavaScript, then I’d also have control over how to transition between the “pages” (or carousel items, if you prefer). There’s currently no way to do that with full page changes.

This is the problem that Jake set out to address in his proposal for navigation transitions a few years back:

Having to reimplement navigation for a simple transition is a bit much, often leading developers to use large frameworks where they could otherwise be avoided. This proposal provides a low-level way to create transitions while maintaining regular browser navigation.

I love this proposal. It focuses on user needs. It also asks why people reach for JavaScript frameworks instead of using what browsers provide. People reach for JavaScript frameworks because browsers don’t yet provide some functionality: components like tabs or accordions; DOM diffing; control over styling complex form elements; navigation transitions. The problems that JavaScript frameworks are solving today should be seen as the R&D departments for web standards of tomorrow. (And conversely, I strongly believe that the aim of any good JavaScript framework should be to make itself redundant.)

I linked to Jake’s excellent proposal in my shitpost saying:

bucketloads of JavaScript wouldn’t be needed if navigation transitions were available in browsers

But then I added—and I almost didn’t—this:

(not portals)

Now you might be asking yourself what Paul said out loud:

Excuse my ignorance but… WTF are portals!?

I replied with a link to the portals proposal and what I thought was an example use case:

Portals are a proposal from Google that would help their AMP use case (it would allow a web page to be pre-rendered, kind of like an iframe).

That was based on my reading of the proposal:

…show another page as an inset, and then activate it to perform a seamless transition to a new state, where the formerly-inset page becomes the top-level document.

It sounded like Google’s top stories carousel. And the proposal goes into a lot of detail around managing cross-origin requests. Again, that strikes me as something that would be more useful for a search engine than a single page app.

But Jake was not happy with my description. I didn’t intend to besmirch portals by mentioning Google AMP in the same sentence, but I can see how the transitive property of ickiness would apply. Because Google AMP is a nasty monopolistic project that harms the web and is an embarrassment to many open web advocates within Google, drawing any kind of comparison to AMP is kind of like Godwin’s Law for web stuff. I know that makes it sounds like I’m comparing Google AMP to Hitler, and just to be clear, I’m not (though I have myself been called a fascist by one of the lead engineers on AMP).

Clearly, emotions run high when Google AMP is involved. I regret summoning its demonic presence.

After chatting with Jake some more, I tried to find a better use case to describe portals. Reading the proposal, portals sound a lot like “spicy iframes”. So here’s a different use case that I ran past Jake: say you’re on a website that has an iframe embedded in it—like a YouTube video, for example. With portals, you’d have the ability to transition the iframe to a fully-fledged page smoothly.

But Jake told me that even though the proposal talks a lot about iframes and cross-origin security, portals are conceptually more like using rel="prerender" …but then having scripting control over how the pre-rendered page becomes the current page.

Put like that, portals sound more like Jake’s original navigation transitions proposal. But I have to say, I never would’ve understood that use case just from reading the portals proposal. I get that the proposal is aimed more at implementators than authors, but in its current form, it doesn’t seem to address the use case of single page apps.

Kenji said:

we haven’t seen interest from SPA folks in portals so far.

I’m not surprised! He goes on:

Maybe, they are happy / benefits aren’t clear yet.

From my own reading of the portals proposal, I think the benefits are definitely not clear. It’s almost like the opposite of Jake’s original proposal for navigation transitions. Whereas as that was grounded in user needs and real-world examples, the portals proposal seems to have jumped to the intricacies of implementation without covering the user needs.

Don’t get me wrong: if portals somehow end up leading to a solution more like Jake’s navigation transitions proposals, then I’m all for that. That’s the end result I care about. I’d love it if people had a lightweight option for getting the perceived benefits of single page apps without the costly overhead in performance that comes with JavaScripting all the things.

I guess the web I want includes giant carousels.

Downloading from Google Fonts

If you’re using web fonts, there are good performance (and privacy) reasons for hosting your own font files. And fortunately, Google Fonts gives you that option. There’s a “Download family” button on every specimen page.

But if you go ahead and download a font family from Google Fonts, you’ll notice something a bit odd. The .zip file only contains .ttf files. You can serve those on the web, but it’s far from the best choice. Woff2 is far leaner in file size.

This means you need to manually convert the downloaded .ttf files into .woff or .woff2 files using something like Font Squirrel’s generator. That’s fine, but I’m curious as to why this step is necessary. Why doesn’t Google Fonts provide .woff or .woff2 files in the downloaded folder? After all, if you choose to use Google Fonts as a third-party hosting service for your fonts, it most definitely serves up the appropriate file formats.

I thought maybe it was something to do with the licensing. Maybe some licenses only allow for unmodified truetype files to be distributed? But I’ve looked at fonts with different licenses—some have Apache 2 licensing, some have Open Font licensing—and they’re all quite permissive and definitely allow for modification.

Maybe the thinking is that, if you’re hosting your own font files, then you know what you’re doing and you should be able to do your own file conversion and subsetting. But I’ve come across more than one website in the wild serving up .ttf files. And who can blame them? They want to host their own font files. They downloaded those files from Google Fonts. Why shouldn’t they assume that they’re good to go?

It’s all a bit strange. If anyone knows why Google Fonts only provides .ttf files for download, please let me know. In a pinch, I will also accept rampant speculation.

Trys also pointed out some weird default behaviour if you do let Google Fonts do the hosting for you. Specifically if it’s a variable font. Let’s say it’s a font with weight as a variable axis. You specify in advance which weights you’ll be using, and then it generates separate font files to serve for each different weight.

Doesn’t that defeat the whole point of using a variable font? I mean, I can see how it could result in smaller file sizes if you’re just using one or two weights, but isn’t half the fun of having a weight axis that you can go crazy with as many weights as you want and it’s all still one font file?

Like I said, it’s all very strange.

Performance and people

I was helping a client with a bit of a performance audit this week. I really, really enjoy this work. It’s such a nice opportunity to get my hands in the soil of a website, so to speak, and suggest changes that will have a measurable effect on the user’s experience.

Not only is web performance a user experience issue, it may well be the user experience issue. Page speed has a proven demonstrable direct effect on user experience (and revenue and customer satisfaction and whatever other metrics you’re using).

It struck me that there’s a continuum of performance challenges. On one end of the continuum, you’ve got technical issues. These can be solved with technical solutions. On the other end of the continuum, you’ve got human issues. These can be solved with discussions, agreement, empathy, and conversations (often dreaded or awkward).

I think that, as developers, we tend to gravitate towards the technical issues. That’s our safe space. But I suspect that bigger gains can be reaped by tackling the uncomfortable human issues.

This week, for example, I uncovered three performance issues. One was definitely technical. One was definitely human. One was halfway between.

The technical issue was with web fonts. It’s a lot of fun to dive into this aspect of web performance because quite often there’s some low-hanging fruit: a relatively simple technical fix that will boost the performance (or perceived performance) of a website. That might be through resource hints (using link rel=“preload” in the HTML) or adjusting the font loading (using font-display in the CSS) or even nerdier stuff like subsetting.

In this case, the issue was with the file format of the font itself. By switching to woff2, there were significant file size savings. And the great thing is that @font-face rules allow you to specify multiple file formats so you can still support older browsers that can’t handle woff2. A win all ‘round!

The performance issue that was right in the middle of the technical/human continuum was with images. At first glance it looked like a similar issue to the fonts. Some images were being served in the wrong formats. When I say “wrong”, I guess I mean inappropriate. A photographic image, for example, is probably going to best served as a JPG rather than a PNG.

But unlike the fonts, the images weren’t in the direct control of the developers. These images were coming from a Content Management System. And while there’s a certain amount of processing you can do on the server, a human still makes the decision about what file format they’re uploading.

I’ve seen this happen at Clearleft. We launched an event site with lean performant code, but then someone uploaded an image that’s megabytes in size. The solution in that case wasn’t technical. We realised there was a knowledge gap around image file formats—which, let’s face it, is kind of a techy topic that most normal people shouldn’t be expected to know.

But it was extremely gratifying to see that people were genuinely interested in knowing a bit more about choosing the right format for the right image. I was able to provide a few rules of thumb and point to free software for converting images. It empowered those people to feel more confident using the Content Management System.

It was a similar situation with the client site I was looking at this week. Nobody is uploading oversized images in order to deliberately make the site slower. They probably don’t realise the difference that image formats can make. By having a discussion and giving them some pointers, they’ll have more knowledge and the site will be faster. Another win all ‘round!

At the other end of continuum was an issue that wasn’t technical. From a technical point of view, there was just one teeny weeny little script. But that little script is Google Tag Manager which then calls many, many other scripts that are not so teeny weeny. Third party scripts …the bane of web performance!

In retrospect, it seems unbelievable that third-party JavaScript is even possible. I mean, putting arbitrary code—that can then inject even more arbitrary code—onto your website? That seems like a security nightmare!

Remember when I did a countdown of the top four web performance challenges? At the number one spot is other people’s JavaScript.

Now one technical solution would be to remove the Google Tag Manager script. But that’s probably not very practical—you’ll probably just piss off some other department. That said, if you can’t find out which department was responsible for adding the Google Tag Manager script in the first place, it might we well be an option to remove it and then wait and see who complains. If no one notices it’s gone, job done!

More realistically, there’s someone who’s added that Google Tag Manager script for their own valid reasons. You’ll need to talk to them and understand their needs.

Again, as with images uploaded in a Content Management System, they may not be aware of the performance problems caused by third-party scripts. You could try throwing numbers at them, but I think you get better results by telling the story of performance.

Use tools like Request Map Generator to help them visualise the impact that third-party scripts are having. Talk to them. More importantly, listen to them. Find out why those scripts are being requested. What are the outcomes they’re working towards? Can you offer an alternative way of providing the data they need?

I think many of us developers are intimidated or apprehensive about approaching people to have those conversations. But it’s necessary. And in its own way, it can be as rewarding as tinkering with code. If the end result is a faster website, then the work is definitely worth doing—whether it’s technical work or people work.

Personally, I just really enjoy working on anything that will end up improving a website’s performance, and by extension, the user experience. If you fancy working with me on your site, you should get in touch with Clearleft.

Hard to break

I keep thinking about some feedback that Cassie received recently.

She had delivered the front-end code for a project at Clearleft, and—this being Cassie we’re talking about—the code was rock solid. The client’s Quality Assurance team came back with the verdict that it was “hard to break.”

Hard to break. I love that. That might be the best summation I’ve heard for describing resilience on the web.

If there’s a corollary to resilient web design, it would be brittle web design. In a piece completely unrelated to web development, Jamais Cascio describes brittle systems:

When something is brittle, it’s susceptible to sudden and catastrophic failure.

That sounds like an inarguably bad thing. So why would anyone end up building something in a brittle way? Jamais Cascio continues:

Things that are brittle look strong, may even be strong, until they hit a breaking point, then everything falls apart.

Ah, there’s the rub! It’s not that brittle sites don’t work. They work just fine …until they don’t.

Brittle systems are solid until they’re not. Brittleness is illusory strength. Things that are brittle are non-resilient, sometimes even anti-resilient — they can make resilience more difficult.

Kilian Valkhof makes the same point when it comes to front-end development. For many, accessibility is an unknown unknown:

When you start out it’s you, notepad and a browser against the world. You open up that notepad, and you type

<div onclick="alert('hello world');">Click me!</div>

You fire up your browser, you click your div and …it works! It just works! Awesome. You open up the devtools. No errors. Well done! Clearly you did a good job. On to the next thing.

At the surface level, there’s no discernable difference between a resilient solution and a brittle one:

For all sorts of reasons, both legitimate and, as always, weird browser legacy reasons, a clickable div will mostly work. Well enough to fool someone starting out anyway.

If everything works, how would they know it kinda doesn’t?

Killian goes on to suggest ways to try to make this kind of hidden brittleness more visible.

Furthermore we could envision a browser that is much stricter when developing.

This something I touched on when I was talking about web performance with Gerry on his podcast:

There’s a disconnect in the process we go through when we’re making something, and then how that thing is experienced when it’s actually on the web, which is dependent on network speeds and processing speeds and stuff.

I spend a lot of time wondering why so many websites are badly built. Sure, there’s a lot can be explained by misaligned priorities. And it could just be an expression of Sturgeon’s Law—90% of websites are crap because 90% of everything is crap. But I’ve also come to realise that even though resilience is the antithesis to brittleness, they both share something in common: they’re invisible.

We have a natural bias towards what’s visible. Being committed to making sure something is beautiful to behold is, in some ways, the easy path to travel. But being committed to making sure something is also hard to break? That takes real dedication.


It’s been fascinating to see how television programmes have adapted to The Situation. It’s like there’s been a weird inversion with the YouTube asthetic. Instead of YouTubers doing their utmost to emulate the look of professional television, now everyone on professional television looks like a YouTuber.

No more lighting or audio technicians. No more studio audiences. Heck, no more studios.

There are some kinds of TV programmes that are showing the strain. A lot of comedy formats just fall flat without the usual production values. But a lot of programmes work just fine. In fact, some of them might be better. Watching Mary Beard present Front Row Late from her house is an absolute delight. It feels more direct and honest without the artiface of a television studio. It kind of makes you wonder whether expensive production costs are really necessary when what you really care about is the content.

All of this is one big belaboured metaphor for websites.

In times of crisis, informational websites sometimes offer a “lite” version. Max has even made an emergency website kit:

The site contains only the bare minimum - no webfonts, no tracking, no unnecessary images. The entire thing should fit in a single HTTP request. It’s basically just a small, ultra-lean blog focused on maximum resilience and accessibility. The Service Worker takes it a step further from there so if you’ve visited the site once, the information is still accessible even if you lose network coverage.

Eric emphasises the importance of performance in his post Get Static:

I’m thinking here of sites for places like health departments (and pretty much all government services), hospitals and clinics, utility services, food delivery and ordering, and I’m sure there are more that haven’t occurred to me.  As much as you possibly can, get it down to static HTML and CSS and maybe a tiny bit of enhancing JS, and pare away every byte you can.

Tom Loosemore offers this advice to teams building new coronavirus services:

  1. Get a 4 year-old Android phone, and use it as your test/demo device.
  2. is your friend.
  3. Full React isn’t your friend if it makes your service slow & inaccessible

Remember: This is for everyone.

Indeed, are usually a paragon of best practices in just about any situation. But they dropped the ball recently, as Matthew attests: is a static site, fetching and displaying remote data. It is also a 100% client-side JavaScript React site. is 238K vs 770K (basics) on load. I’ve removed about 550K of JavaScript. It seems to work the same.

As Tom says:

One sign that your website isn’t meeting the needs of all your users is when Matthew Somerville gets sufficiently grumpy about it to do a proper version himself.

It’s true enough that Matthew excels at creating lightweight, accessible versions of services that are too bloated or buggy to use. His accessible Odeon project from back in the day is legendary. And I use his slimline version of the National Rail website all the time:—it’s a terrificly performant progressive web app.

It’s thankless work though. It flies in the face of everything considered “modern” web development. (If you want to know the cost of “modern” framework-driven JavaScript-first web development, Tim has the numbers.) But Matthew is kind of a hero to me. I wish more developers would follow his example.

Maybe now, with this rush to make lightweight versions of valuable services, we might stop and reflect on whether we ever really needed all those added extras in the first place.

Hope springs eternal.

Update: Matthew has written about his process in Looking at

Overlay gap

I think a lot about Danielle’s talk at Patterns Day last year.

Around about the six minute mark she starts talking about gaps and overlaps.

Gaps are where hidden complexity live. If we don’t have a category to cover it, in effect it becomes invisible. But that doesn’t mean it’s not there. Unidentified gaps cause inconsistency and confusion.

Overlaps occur when two separate categories encompass some of the same areas of responsibility. They cause conflict, duplication of effort, and unnecessary friction.

This is the bit I keep thinking about. It’s such an insightful lens to view things through. On just about any project, tensions are almost due to either gaps (“I thought someone else was doing that”) or overlaps (“Oh, you’re doing that? I thought we were doing that”).

When I was talking to Gerry on his new podcast recently, we were trying to figure out why web performance is in such a woeful state. I mused that there may be a gap. Perhaps designers think it’s a technical problem and developers think it’s a design problem. I guess you could try to bridge this gap by having someone whose job is to focus entirely on performance. But I suspect the better—but harder—solution is to create a shared culture of performance, of the kind Lara wrote about in her book:

Performance is truly everyone’s responsibility. Anyone who affects the user experience of a site has a relationship to how it performs. While it’s possible for you to single-handedly build and maintain an incredibly fast experience, you’d be constantly fighting an uphill battle when other contributors touch the site and make changes, or as the Web continues to evolve.

I suspect there’s a similar ownership gap at play when it comes to the ubiquitous obtrusive overlays that are plastered on so many websites these days.

Kirill Grouchnikov recently published a gallery of screenshots showcasing the beauty of modern mobile websites:

There are two things common between the websites in these screenshots that I took yesterday.

  1. They are beautifully designed, with great typography, clear branding, all optimized for readability.
  2. I had to install Firefox, Adblock Plus and uBlock Origin, as well as manually select and remove additional elements such as subscription overlays.

The web can be beautiful. Except it’s not right now.

How is this dissonance possible? How can designers and developers who clearly care about the user experience be responsible for unleashing such user-hostile interfaces?

PM/Legal/Marketing made me do it

I get that. But surely the solution can’t be to shrug our shoulders, pass the buck, and say “not my job.” Somebody designed each one of those obtrusive overlays. Somebody coded up each one and pushed them into production.

It’s clear that this is a problem of communication and understanding, rather than a technical problem. As always. We like to talk about how hard and complex our technical work is, but frankly, it’s a lot easier to get a computer to do what you want than to convince a human. Not least because you also need to understand what that other human wants. As Danielle says:

Recognising the gaps and overlaps is only half the battle. If we apply tools to a people problem, we will only end up moving the problem somewhere else.

Some issues can be solved with better tools or better processes. In most of our workplaces, we tend to reach for tools and processes by default, because they feel easier to implement. But as often as not, it’s not a technology problem. It’s a people problem. And the solution actually involves communication skills, or effective dialogue.

So let’s say it is someone in the marketing department who is pushing to have an obtrusive newsletter sign-up form get shoved in the user’s face. Talk to them. Figure out what their goals are—what outcome are they hoping to get to. If they don’t seem to understand the user-experience implications, talk to them about that. But it needs to be a two-way conversation. You need to understand what they need before you start telling them what you want.

I realise that makes it sound patronisingly simple, and I know that in actuality it’s a sisyphean task. It may be that genuine understanding between people is the wickedest of design problems. But even if this problem seems insurmoutable, at least you’d be tackling the right problem.

Because the web can’t survive like this.

Oh, embed!

I wrote yesterday about how messing about on your own website can be a welcome distraction. I did some tinkering with on the weekend that you might be interested in.

Let me set the scene…

I’ve started recording and publishing a tune a day. I grab my mandolin, open up Quicktime and make a movie of me playing a jig, a reel, or some other type of Irish tune. I include a link to that tune on The Session and a screenshot of the sheet music for anyone who wants to play along. And I embed the short movie clip that I’ve uploaded to YouTube.

Now it’s not the first time I’ve embedded YouTube videos into my site. But with the increased frequency of posting a tune a day, the front page of ended up with multiple embeds. That is not good for performance—my Lighthouse score took quite a hit. Worst of all, if a visitor doesn’t end up playing an embedded video, all of the markup, CSS, and JavaScript in the embedded iframe has been delivered for nothing.

Meanwhile over on The Session, I’ve got a strategy for embedding YouTube videos that’s better for performance. Whenever somebody posts a link to a video on YouTube, the thumbnail of the video is embedded. Only when you click the thumbnail does that image get swapped out for the iframe with the video.

That’s what I needed to do here on

First off, I should explain how I’m embedding things generally ‘round here. Whenever I post a link or a note that has a URL in it, I run that URL through a little PHP script called getEmbedCode.php.

That code checks to see if the URL is from a service that provides an oEmbed endpoint. A what-Embed? oEmbed!

oEmbed is like a minimum viable read-only API. It was specced out by Leah and friends years back. You ping a URL like this:

In this case is the endpoint and url is the value of a URL from that provider. Here’s a real life example from YouTube:

So is the endpoint and url is the address of any video on YouTube.

You get back some JSON with a pre-defined list of values like title and html. That html payload is the markup for your embed code.

By default, YouTube sends back markup like this:

allow="accelerometer; autoplay; encrypted-media; gyroscope; picture-in-picture"

But now I want to use an img instead of an iframe. One of the other values returned is thumbnail_url. That’s the URL of a thumbnail image that looks something like this:

In fact, once you know the ID of a YouTube video (the ?v= bit in a YouTube URL), you can figure out the path to multiple images of different sizes:

(Although that last one—maxresdefault.jpg—might not work for older videos.)

Okay, so I need to extract the ID from the YouTube URL. Here’s the PHP I use to do that:

parse_str(parse_url($url, PHP_URL_QUERY), $arguments);
$id = $arguments['v'];

Then I can put together some HTML like this:

<a class="videoimglink" href="'.$url.'">
<img width="100%" loading="lazy"
srcset="'.$id.'/mqdefault.jpg 320w,'.$id.'/hqdefault.jpg 480w,'.$id.'/maxresdefault.jpg 1280w

Now I’ve got a clickable responsive image that links through to the video on YouTube. Time to enhance. I’m going to add a smidgen of JavaScript to listen for a click on that link.

Over on The Session, I’m using addEventListener but here on I’m going to be dirty and listen for the event directly in the markup using the onclick attribute.

When the link is clicked, I nuke the link and the image using innerHTML. This injects an iframe where the link used to be (by updating the innerHTML value of the link’s parentNode).

this.parentNode.innerHTML='<iframe src='.$id.'?autoplay=1></iframe>'"

But notice that I’m not using the default YouTube URL for the iframe. That would be:

Instead I’m swapping out the domain for

I can’t remember where I first came across this undocumented parallel version of YouTube that has, yes, you guessed it, no cookies. It turns out that, not only is the default YouTube embed code bad for performance, it is—unsurprisingly—bad for privacy too. So the domain can protect your site’s visitors from intrusive tracking. Pass it on.

Anyway, I’ve got the markup I want now:

<a class="videoimglink" href=""
this.parentNode.innerHTML='<iframe src=></iframe>'">
<img width="100%" loading="lazy"
alt="The Banks Of Lough Gowna (jig) on mandolin"
srcset=" 320w, 480w, 1280w

The functionality is all there. But I want to style the embedded images to look more like playable videos. Time to break out some CSS (this is why I added the videoimglink class to the YouTube link).

.videoimglink {
    display: block;
    position: relative;

I’m going to use generated content to create a play button icon. Because I can’t use generated content on an img element, I’m applying these styles to the containing .videoimglink a element.

.videoimglink::before {
    content: '▶';

I was going to make an SVG but then I realised I could just be lazy and use the unicode character instead.

Right. Time to draw the rest of the fucking owl:

.videoimglink::before {
    content: '▶';
    display: inline-block;
    position: absolute;
    background-color: var(--background-color);
    color: var(--link-color);
    border-radius: 50%;
    width: 10vmax;
    height: 10vmax;
    top: calc(50% - 5vmax);
    left: calc(50% - 5vmax);
    font-size: 6vmax;
    text-align: center;
    text-indent: 1vmax;
    opacity: 0.5;

That’s a bunch of instructions for sizing and positioning. I’d explain it, but that would require me to understand it and frankly, I’m not entirely sure I do. But it works. I think.

With a translucent play icon positioned over the thumbnail, all that’s left is to add a :hover style to adjust the opacity:

.videoimglink:focus::before {
    opacity: 0.75;

Wheresoever thou useth :hover, thou shalt also useth :focus.

Okay. It’s good enough. Ship it!

The Banks Of Lough Gowna (jig) on mandolin

If you embed YouTube videos on your site, and you’d like to make them more performant, check out this custom element that Paul made: Lite YouTube Embed. And here’s a clever technique that uses the srcdoc attribute to get a similar result (but don’t forget to use the domain).

Lighthouse bookmarklet

I use Firefox. You should too. It’s fast, secure, and more privacy-focused than the leading browser from the big G.

When it comes to web development, the CSS developer tooling in Firefox is second-to-none. But when it comes to JavaScript and network-related debugging (like service workers), Chrome’s tools are currently better than Firefox’s (for now). For example, Chrome has a tab in its developer tools that lets you run Lighthouse on the currently open tab.

Yesterday, I got the Calibre newsletter, which always has handy performance-related links from Karolina. She pointed to a Lighthouse extension for Firefox. “Excellent!”, I thought, and I immediately installed it. But I had some qualms about installing a plug-in from Google into a browser from Mozilla, particularly as the plug-in page says:

This is not a Recommended Extension. Make sure you trust it before installing

Well, I gave it a go. It turns out that all it actually does is redirect to the online version of Lighthouse. “Hang on”, I thought. “This could just be a bookmarklet!”

So I immediately uninstalled the browser extension and made this bookmarklet:


Drag that up to your desktop browser’s bookmarks toolbar. Press it whenever you’re on a site that you want to test.

Telling the story of performance

At Clearleft, we’ve worked with quite a few clients on site redesigns. It’s always a fascinating process, particularly in the discovery phase. There’s that excitement of figuring out what’s currently working, what’s not working, and what’s missing completely.

The bulk of this early research phase is spent diving into the current offering. But it’s also the perfect time to do some competitor analysis—especially if we want some answers to the “what’s missing?” question.

It’s not all about missing features though. Execution is equally important. Our clients want to know how their users’ experience shapes up compared to the competition. And when it comes to user experience, performance is a huge factor. As Andy says, performance is a UX problem.

There’s no shortage of great tools out there for measuring (and monitoring) performance metrics, but they’re mostly aimed at developers. Quite rightly. Developers are the ones who can solve most performance issues. But that does make the tools somewhat impenetrable if you don’t speak the language of “time to first byte” and “first contentful paint”.

When we’re trying to show our clients the performance of their site—or their competitors—we need to tell a story.

Web Page Test is a terrific tool for measuring performance. It can also be used as a story-telling tool.

You can go to if you don’t need to tweak settings much beyond the typical site visit (slow 3G on mobile). Pop in your client’s URL and, when the test is done, you get a valuable but impenetrable waterfall chart. It’s not exactly the kind of thing I’d want to present to a client.

Fortunately there’s an attention-grabbing output from each test: video. Download the video of your client’s site loading. Then repeat the test with the URL of a competitor. Download that video too. Repeat for as many competitor URLs as you think appropriate.

Now take those videos and play them side by side. Presentation software like Keynote is perfect for showing multiple videos like this.

This is so much more effective than showing a table of numbers! Clients get to really feel the performance difference between their site and their competitors.

Running all those tests can take time though. But there are some other tools out there that can give a quick dose of performance information.

SpeedCurve recently unveiled Page Speed Benchmarks. You can compare the performance of sites within a particualar sector like travel, retail, or finance. By default, you’ll get a filmstrip view of all the sites loading side by side. Click through on each one and you can get the video too. It might take a little while to gather all those videos, but it’s quicker than using Web Page Test directly. And it might be that the filmstrip view is impactful enough for telling your performance story.

If, during your discovery phase, you find that performance is being badly affected by third-party scripts, you’ll need some way to communicate that. Request Map Generator is fantastic for telling that story in a striking visual way. Pop the URL in there and then take a screenshot of the resulting visualisation.

The beginning of a redesign project is also the time to take stock of current performance metrics so that you can compare the numbers after your redesign launches. is really great for tracking performance over time. You won’t get any videos but you will get some very appealing charts and graphs.

Web Page Test, Page Speed Benchmarks, and Request Map Generator are great for telling the story of what’s happening with performance right balances that with the story of performance over time.

Measuring performance is important. Communicating the story of performance is equally important.


As you may have noticed, I’m a fan of progressive enhancement.

It’s not cool. It’s often at odds with “modern” web development, so I end up looking like an old man yelling at a cloud to get off my lawn. Or something.

At its heart though, progressive enhancement seems fairly uncontroversial and inoffensive to me. It’s an approach. A mindset. Here’s how I describe it in Resilient Web Design:

  1. Identify core functionality.
  2. Make that functionality available using the simplest possible technology.
  3. Enhance!

Progressive enhancement makes use of the principle of least power:

Choose the least powerful language suitable for a given purpose.

That’s step two of the three-step process. But the third step is vital.

I think a lot of the hostility towards progressive enhancement comes from a misunderstanding of that three-step process, perhaps thinking that it stops at step two. I’m sure that some have intrepreted progressive enhancement as preventing developers from using the latest and greatest technology. Nothing could be further from the truth!

Taking a layered approach to building on the web gives you permission to try cutting‐edge JavaScript APIs, regardless of how many or how few browsers currently implement them.

The most common misunderstanding of progressive enhancement is that it’s inherently about JavaScript. That’s not true. You can apply progressive enhancement at every step of front-end development: HTML, CSS, and JavaScript.

But because of JavaScript’s strict error-handling model (at least compared to HTML and CSS), it’s in the JavaScript layer that the lack of a progressive enhancement mindset is most often felt.

That’s why I was saddened by the rise of frameworks and mindsets that assume the availability of JavaScript. Single page apps generally follow this assumption. Everything is delivered via JavaScript: content, markup, styles, and behaviour.

This leads to a terrible situation for performance. The user is left staring at a blank screen, waiting for something—anything!—to appear. Browsers are optimised to stream HTML as soon as they can. Delivering your content via JavaScript rather than HTML means you’re not taking advantage of that optimisation. Your users suffer.

But I was very heartened when I saw the pendulum start to swing back the other way a bit…

Let’s say you’re using a JavaScript framework like React. But the reason you’re using it isn’t because you’re doing anything particularly complex in the browser involving state management. You might be using React because you really like the way it encourages modularity and componentisation.

A few years ago, making a single page app was pretty much the only way you could use React. For you as a developer to experience the benefits of modularity and componentisation, users had to pay the price in the payload (and fragility) of client-side JavaScript.

That’s no longer the case. Now that we can run JavaScript on the server, it’s possible to build in a modular, componentised way and still use progressive enhancement.

When I first heard about Gatsby and Next.js, I thought that was the selling point. Run React on the server; send pre-generated HTML down the wire to the user; then enhance with client-side JavaScript.

But that’s not exactly how it works. The pre-generated HTML isn’t functional. It still needs a bucketload of JavaScript before it can do anything. The actual process is: Run React on the server; send pre-generated HTML down the wire to the user; then send everything again but this time in JavaScript, bundled with the entire React library.

This leads to a situation for users that’s almost worse than before. Instead of staring at a blank screen, now they get HTML lickety-split—excellent! But if they try to interact with what’s on screen, they’ll find that nothing is working yet. Even worse, once the JavaScript is delivered, and is being parsed, they probably can’t even scroll—their device is too busy interpreting all that JavaScript. Your users suffer.

All your content is sent twice. First HTML is sent from the server. These days this is called “server-side rendering”, even though for decades the technical term was “serving a web page” (I’m pretty sure the rendering part happens in a browser). Then a JavaScript library—plus all your bespoke JavaScript—is loaded. Then all your content is loaded again as JSON.

So you’ve got a facade of an interface that you can’t actually interact with until a deluge of JavaScript has been loaded, parsed and executed. The term used for this stage of the process is “hydration”, which makes it sound more like a relaxing treatment from Gwyneth Paltrow than the horrible user experience it is.

The idea is that subsequent navigations—which will happen with Ajax—should be snappy. But the price has already been paid by then. The initial loading experience is jagged and frustrating.

Don’t get me wrong: server-side rendering is great …if what you’re sending from the server is functional. It’s the combination of hollow HTML sent from the server, followed by a huge browser-freezing dump of JavaScript that is an anti-pattern.

This use of server-side rendering followed by hydration feels like progressive enhancement, because it separates out the delivery of markup and scripts. But it’s missing the mindset.

The layered approach of progressive enhancement echoes the separation of concerns in the front-end stack: HTML, CSS, and JavaScript—each layer expressing more power. But while these concepts are related, they’re not interchangable. Separating out the layers of your tech stack isn’t necessarily progressive enhancement. If you have some HTML that relies on JavaScript to be useful, then there’s no benefit in separating that HTML into a separate payload. The HTML that you initially send down the wire needs to be functional (at least at a basic level) before the JavaScript arrives.

I was a little disappointed to see Kyle Simpson—who I admire greatly—conflate separation of concerns with progressive enhancement in his talk from JSCamp 2019:

This content is here. I can see it, and it’s even styled. But I can’t click on the damn button because nothing has loaded in the JavaScript layer yet.

Anybody experienced that where you’ve been on a web page and it’s not really fully functional yet? I can see something but I can’t actually make any usage of it yet.

These are all things that cropped out of our thought process that said: “Let’s build the web in layers. Let’s deliver it progressively in layers. Because that’s morally right. We call this progressive enhancement. And let’s not worry too much about all these potential user experience flaws that may happen.”

That’s a spot-on description of server-side rendering and hydration, but it’s a gross mischaracterisation of progressive enhancement.

That button that requires JavaScript to work? That should’ve been generated with JavaScript. (For example, if you’re building a complex web app, consider sending a read-only view down the wire in HTML—then add any interactive interface elements with JavaScript in the browser.)

If people are equating progressive enhancement with thoughtless server-side rendering and hydration, then I can see why they’d be hostile towards it.

Users would be better served with unprogressive non-enhancement:

You take some structured content, which follows the vertical flow of the document in a way that everyone understands.

Which people traverse easily by either dragging their scroll bar with their mouse, or operating the keyboard using the up and down keys, or using the spacebar.

Or if they’re using a touch device, simply flicking backwards and forwards in that easy way that we’ve all become used to. What you do is you take that, and you fucking well leave it alone.

Alas, that’s not what tools like Gatsby offer. The latest post on their blog is called Why Gatsby is better with JavaScript:

But what about sites or pages where there is no client-side interactivity? Even for those pages, Gatsby offers performance benefits by including JavaScript.

I beg to differ.

(By the way, that same blog post also initially tried to equate the performance hit of client-side JavaScript with the performance hit of images. Andy explains why that’s disingenuous.)

Hope is on the horizon for React in the form of partial hydration. I sincerely hope that it will become the default way of balancing server-side rendering with just-in-time client-side interaction.

The situation we have now is the worst of both worlds: server-side rendering followed by a tsunami of hydration. It has a whiff of progressive enhancement to it (because there’s a cosmetic separation of concerns) but it has none of the user benefits.

Third party

The web turned 30 this year. When I was back at CERN to mark this anniversary, there was a lot of introspection and questioning the direction that the web has taken. Everyone I know that uses the web is in agreement that tracking and surveillance are out of control. It seems only right to question whether the web has lost its way.

But here’s the thing: the technologies that enable tracking and surveillance didn’t exist in the early years of the web—JavaScript and cookies.

Without cookies, the web was stateless. This was by design. Now, I totally understand why cookies—or something like cookies—were needed. Without some way of keeping track of state, there’s no good way for a website to “remember” what’s in your shopping cart, or whether you’ve authenticated yourself.

But why would cookies ever need to work across domains? Authentication, shopping carts and all that good stuff can happen on the same domain. Third-party cookies, on the other hand, seem custom made for tracking and frankly, not much else.

Browsers allow you to disable third-party cookies, though it’s not yet the default. If enough people do it—and complain about the sites that stop working when third-party cookies are disabled—then maybe it can become the default.

Firefox is taking steps in this direction, automatically disabling some third-party cookies—the ones that known trackers. Safari is also taking steps to prevent cross-site tracking. It’s not too late to change the tide of third-party cookies.

Then there’s third-party JavaScript.

In retrospect, it seems unbelievable that third-party JavaScript is even possible. I mean, putting arbitrary code—that can then inject even more arbitrary code—onto your website? That seems like a security nightmare!

I imagine if JavaScript were being specced today, it would almost certainly be restricted to the same origin by default. But I guess the precedent had been set with images and style sheets: they could be embedded regardless of whether their domain names matched yours. Still, this is executable code we’re talking about here: that’s quite a footgun that the web has given site owners. And boy, oh boy, has it been used by the worst people to do the most damage.

Again, as with cookies, if we were to imagine what the web would be like if JavaScript was restricted by a same-domain policy, there are certainly things that would be trickier to do.

  • Embedding video, audio, and maps would get a lot finickier.
  • Analytics would need to be self-hosted. I don’t think that would bother any site owners. An analytics platform like Google Analytics that tracks people across domains is doing it for its own benefit rather than that of site owners.
  • Advertising wouldn’t be creepy and annoying. Instead of what’s so euphemistically called “personalisation”, advertisers would have to rely on serving relevant ads based on the content of the site rather than an invasive psychological profile of the user. (I honestly think that advertisers would benefit from this kind of targetting.)

It’s harder to imagine putting the genie back in the bottle when it comes to third-party JavaScript than it is with third-party cookies. All the same, I wish that browsers made it easier to experiment with it. Just as I can choose to accept all cookies, reject all cookies, or only accept same-origin cookies, I wish I could accept all JavaScript, reject all JavaScript, or only accept same-origin JavaScript.

As it is, browsers are making it harder and harder to exercise any control over JavaScript at all. So we reach for third-party tools. We don’t call them JavaScript managers though. We call them ad blockers. But honestly, most of the ad-blocker users I know—myself included—are not bothered by the advertising; we’re bothered by the tracking. We should really call them surveillance blockers.

If third-party JavaScript weren’t the norm, not only would it make the web more secure, it would make it way more performant. Read the chapter on third parties in this year’s newly-released Web Almanac. The figures are staggering.

93% of pages include at least one third-party resource, 76% of pages issue a request to an analytics domain, the median page requests content from at least 9 unique third-party domains that represent 35% of their total network activity, and the most active 10% of pages issue a whopping 175 third-party requests or more.

I don’t think all the web’s performance ills are due to third-party scripts; developers are doing a bang-up job of making their sites big and bloated with their own self-hosted frameworks and code. But as long as third-party JavaScript is allowed onto a site, there’s a limit to how much good developers can do to improve the performance of their sites.

I go to performance-related conferences and you know who I’ve never seen at those events? The people who write the JavaScript for third-party tracking scripts. Those developers are wielding an outsized influence on the health of the web.

I’m very happy to see the work being done by Mozilla and Apple to normalise the idea of rejecting third-party cookies. I’d love to see the rejection of third-party JavaScript normalised in the same way. I know that it would make my life as a developer harder. But that’s of lesser importance. It would be better for the web.

FF Conf 2019

Friday was FF Conf day here in Brighton. This was the eleventh(!) time that Remy and Julie have put on the event. It was, as ever, excellent.

It’s a conference that ticks all the boxes for me. For starters, it’s a single-track event. The more I attend conferences, the more convinced I am that multi-track events are a terrible waste of time for attendees (and a financially bad model for organisers). I know that sounds like a sweeping broad generalisation, but ask me about it next time we meet and I’ll go into more detail. For now, I just want to talk about this mercifully single-track conference.

FF Conf has built up a rock-solid reputation over the years. I think that’s down to how Remy curates it. He thinks about what he wants to know and learn more about, and then thinks about who to invite to speak on those topics. So every year is like a snapshot of Remy’s brain. By happy coincidence, a snapshot of Remy’s brain right now looks a lot like my own.

You could tell that Remy had grouped the talks together in themes. There was a performance-themed chunk right after lunch. There was a people-themed chunk in the morning. There was a creative-coding chunk at the end of the day. Nice work, DJ.

I think it was quite telling what wasn’t on the line-up. There were no talks about specific libraries or frameworks. For me, that was a blessed relief. The only technology-specific talk was Alice’s excellent talk on Git—a tool that’s useful no matter what you’re coding.

One of the reasons why I enjoyed the framework-free nature of the day is that most talks—and conferences—that revolve around libraries and frameworks are invariably focused on the developer experience. Think about it: next time you’re watching a talk about a framework or library, ask yourself how it impacts user experience.

At FF Conf, the focus was firmly on people. In the case of Laura’s barnstorming presentation, those people are end users (I’m constantly impressed by how calm and measured Laura remains even when talking about blood-boilingly bad behaviour from the tech industry). In the case of Amina’s talk, the people are junior developers. And for Sharon’s presentation, the people are everyone.

One of the most useful talks of the day was from Anna who took us on a guided tour of dev tools to identify performance improvements. I found it inspiring in a very literal sense—if I had my laptop with me, I think I would’ve opened it up there and then and started tinkering with my websites.

Harry also talked about performance, but at Remy’s request, it was more business focused. Specifically, it was focused on Harry’s consultancy business. I think this would’ve been the perfect talk for more of an “industry” event, whereas FF Conf is very much a community event: Harry’s semi-serious jibes about keeping his performance secrets under wraps didn’t quite match the generous tone of the rest of the line-up.

The final two talks from Charlotte and Suz were a perfect double whammy.

When I saw Charlotte speak at Material in Iceland last year, I wrote this aside in my blog post summary:

(Oh, and Remy, when you start to put together the line-up for next year’s FF Conf, be sure to check out Charlotte Dann—her talk at Material was the perfect mix of code and creativity.)

I don’t think I can take credit for Charlotte being on the line-up, but I will take credit for saying she’d be the perfect fit.

And then Suz Hinton closed out the conference with this rallying cry that resonated perfectly with Laura’s talk:

Less mass-produced surveillance bullshit and more Harry Potter magic (please)!

I think that rallying cry could apply equally well to conferences, and I think FF Conf is a good example of that ethos in action.

Request mapping

The Request Map Generator is a terrific tool. It’s made by Simon Hearne and uses the WebPageTest API.

You pop in a URL, it fetches the page and maps out all the subsequent requests in a nifty interactive diagram of circles, showing how many requests third-party scripts are themselves generating. I’ve found it to be a very effective way of showing the impact of third-party scripts to people who aren’t interested in looking at waterfall diagrams.

I was wondering… Wouldn’t it be great if this were built into browsers?

We already have a “Network” tab in our developer tools. The purpose of this tab is to show requests coming in. The browser already has all the information it needs to make a diagram of requests in the same that the request map generator does.

In Firefox, there’s a little clock icon in the bottom left corner of the “Network” tab. Clicking that shows a pie-chart view of requests. That’s useful, but I’d love it if there were the option to also see the connected circles that the request map generator shows.

Just a thought.

The Weight of the WWWorld is Up to Us by Patty Toland

It’s Patty Toland’s first time at An Event Apart! She’s from the fantabulous Filament Group. They’re dedicated to making the web work for everyone.

A few years ago, a good friend of Patty’s had a medical diagnosis that required everyone to pull together. Another friend shared an article about how not to say the wrong thing. This is ring theory. In a moment of crisis, the person involved is in the centre. You need to understand where you are in this ring structure, and only ever help and comfort inwards and dump concerns and problems outwards.

At the same time, Patty spent time with her family at the beach. Everyone reads the same books together. There was a book about a platoon leader in Vietnam. 80% of the story was literally a litany of stuff—what everyone was carrying. This was peppered with the psychic and emotional loads that they were carrying.

A month later there was a lot of coverage of Syrian refugees arriving in Europe. People were outraged to see refugees carrying smartphones as though that somehow showed they weren’t in a desperate situation. But smartphones are absolutely a necessity in that situation, and most of the phones were less expensive, lower-end devices. was a useful site for people in crisis, but the navigation was designed to require JavaScript.

When people thing about mobile, they think about freedom and mobility. But with that JavaScript decision, the developers piled baggage on to the users.

There was a common assertion that slow networks were a third-world challenge. Remember Facebook’s network challenges? They always talked about new markets in India and Africa. The implication is that this isn’t our problem in, say, Omaha or New York.

Pew Research provided a lot of data back then that showed that this thinking was wrong. Use of cell phones, especially smartphones and tablets, escalated dramatically in the United States. There was a trend towards mobile-only usage. This was in low-income households—about one third of the population. Among 5,400 panelists, 15% did not have a JavaScript-enabled device.

Pew Research provided updated data this year. The research shows an increase in those trends. Half of the population access the web primarily on mobile. The cost of a broadband subscription is too expensive for many people. Sometimes broadband access simply isn’t available.

There’s a term called “the homework gap.” Two thirds of teachers assign broadband-dependent homework, while one third of students have no access to broadband.

At most 37% of people have unlimited data. Most people run out of data on a frequent basis.

Speed also varies wildly. 4G doesn’t really mean anything. The data is all over the place.

This shows that network issues are definitely not just a third world challenge.

On the 25th anniversary of the web, Tim Berners-Lee said the web’s potential was only just beginning to be glimpsed. Everyone has a role to play to ensure that the web serves all of humanity. In his contract for the web, Tim outlined what governments, companies, and users need to do. This reminded Patty of ring theory. The user is at the centre. Designers and developers are in the next circle out. Then there’s the circle of companies. Then there are platforms, browsers, and frameworks. Finally there’s the outer circle of governments.

Are we helping in or dumping in? If you look at the data for the average web page size (2 megabytes), we are definitely dumping in. The size of third-party JavaScript has octupled.

There’s no way for a user to know before clicking a link how big and bloated the page is going to be. Even if they abandon the page load, they’ve still used (and wasted) a lot of data.

Third party scripts—like ads—are really bad at dumping in (to use the ring theory model). The best practices for ads suggest that up to 100 additional HTTP requests is totally acceptable. Unbelievable! It doesn’t matter how performant you’ve made a site when this crap gets piled on top of it.

In 2018, the internet’s data centres alone may already have had the same carbon footprint as all global air travel. This will probably triple in the next seven years. The amount of carbon it takes to train a single AI algorithm is more than the entire life cycle of a car. Then there’s fucking Bitcoin. A single Bitcoin transaction could power 21 US households. It is designed to use—specifically, waste—more and more energy over time.

What should we be doing?

Accessibility should be at the heart of what we build. Plan, test, educate, and advocate. If advocacy doesn’t work, fear can be a motivator. There’s an increase in accessibility lawsuits.

Our websites should be as light as possible. Ask, measure, monitor, and optimise. RequestMap is a great tool for visualising requests. You can see the size and scale of third-party requests. You can also see when images are far, far bigger than they need to be.

Take a critical guide to everything and pare everything down. Set perforance budgets—file size budgets, for example. Optimise images, subset custom fonts, lazyload images and videos, get third-party tools out of the critical path (or out completely), and seek out lighter frameworks.

Test on real devices that real people are using. See Alex Russell’s data on the differences between the kind of devices we use and typical low-end devices. We literally need to stop people in JavaScript.

Push the boundaries. See the amazing work that Adrian Holovaty did with Soundslice. He had to make on-the-fly sheet music generation work on old iPads that musicians like to use. He recommends keeping old devices around to see how poorly your product is working on it.

If you have some power, then your job is to empower somebody else.

—Toni Morrison

Opening up the AMP cache

I have a proposal that I think might alleviate some of the animosity around Google AMP. You can jump straight to the proposal or get some of the back story first…

The AMP format

Google AMP is exactly the kind of framework I’d like to get behind. Unlike most front-end frameworks, its components take a declarative approach—no knowledge of JavaScript required. I think Lea’s excellent Mavo is the only other major framework that takes this inclusive approach. All the configuration happens in markup, and all the styling happens in CSS. Excellent!

But I cannot get behind AMP.

Instead of competing on its own merits, AMP is unfairly propped up by the search engine of its parent company, Google. That makes it very hard to evaluate whether AMP is being used on its own merits. Instead, the evidence suggests that most publishers of AMP pages are doing so because they feel they have to, rather than because they want to. That’s a real shame, because as a library of web components, AMP seems pretty good. But there’s just no way to evaluate AMP-the-format without taking into account AMP-the-ecosystem.

The AMP ecosystem

Google AMP ostensibly exists to make the web faster. Initially the focus was specifically on mobile performance, but that distinction has since fallen by the wayside. The idea is that by using AMP’s web components, your pages will be speedy. Though, as Andy Davies points out, this isn’t always the case:

This is where I get confused… only have an AMP site yet it’s performance is awful from a user perspective - isn’t AMP supposed to prevent this?

See also: Google AMP lowered our page speed, and there’s no choice but to use it:

According to Google’s own Page Speed Insights audit (which Google recommends to check your performance), the AMP version of articles got an average performance score of 87. The non-AMP versions? 95.

Publishers who already have fast web pages—like The Guardian—are still compelled to make AMP versions of their stories because of the search benefits reserved for AMP. As Terence Eden reported from a meeting of the AMP advisory committee:

We heard, several times, that publishers don’t like AMP. They feel forced to use it because otherwise they don’t get into Google’s news carousel — right at the top of the search results.

Some people felt aggrieved that all the hard work they’d done to speed up their sites was for nothing.

The Google AMP team are at pains to point out that AMP is not a ranking factor in search. That’s true. But it is unfairly privileged in other ways. Only AMP pages can appear in the Top Stories carousel …which appears above any other search results. As I’ve said before:

Now, if you were to ask any right-thinking person whether they think having their page appear right at the top of a list of search results would be considered preferential treatment, I think they would say hell, yes! This is the only reason why The Guardian, for instance, even have AMP versions of their content—it’s not for the performance benefits (their non-AMP pages are faster); it’s for that prime real estate in the carousel.

From A letter about Google AMP:

Content that “opts in” to AMP and the associated hosting within Google’s domain is granted preferential search promotion, including (for news articles) a position above all other results.

That’s not the only way that AMP pages get preferential treatment. It turns out that the secret to the speed of AMP pages isn’t the web components. It’s the prerendering.

The AMP cache

If you’ve ever seen an AMP page in a list of search results, you’ll have noticed the little lightning icon. If you’ve ever tapped on that search result, you’ll have noticed that the page loads blazingly fast!

That’s not down to AMP-the-format, alas. That’s down to the fact that the page has been prerendered by Google before you even went to it. If any page were prerendered that way, it would load blazingly fast. But currently, this privilege is reserved for AMP pages only.

If, after tapping through to that AMP page, you looked at the address bar of your browser, you might have noticed something odd. Even though you might have thought you were visiting The Washington Post, or The New York Times, the URL of the (blazingly fast) page you’re looking at is still under Google’s domain. That’s because Google hosts any AMP pages that it prerenders.

Google calls this “the AMP cache”, but it would be better described as “AMP hosting”. The web page sent down the wire is hosted on Google’s domain.

Here’s that AMP letter again:

When a user navigates from Google to a piece of content Google has recommended, they are, unwittingly, remaining within Google’s ecosystem.

Through gritted teeth, I will refer to this as “the AMP cache”, because that’s what everyone else calls it. But make no mistake, Google is hosting—not caching—these pages.

But why host the pages on a Google domain? Why not prerender the original URLs?

Prerendering and privacy

Scott summed up the situation with AMP nicely:

The pitch I think site owners are hearing is: let us host your pages on our domain and we’ll promote them in search results AND preload them so they feel “instant.” To opt-in, build pages using this component syntax.

But perhaps we could de-couple the AMP format from the AMP cache.

That’s what Terence suggests:

My recommendation is that Google stop requiring that organisations use Google’s proprietary mark-up in order to benefit from Google’s promotion.

The AMP letter, too:

Instead of granting premium placement in search results only to AMP, provide the same perks to all pages that meet an objective, neutral performance criterion such as Speed Index.

Scott reiterates:

It’s been said before but it would be so good for the web if pages with a Lighthouse score over say, 90 could get into that top search result area, even if they’re not built using Google’s AMP framework. Feels wrong to have to rebuild/reproduce an already-fast site just for SEO.

This was also what I was calling for. But then Malte pointed out something that stumped me. Privacy.

Here’s the problem…

Let’s say Google do indeed prerender already-fast pages when they’re listed in search results. You, a search user, type something into Google. A list of results come back. Google begins pre-rendering some of them. But you don’t end up clicking through to those pages. Nonetheless, the servers those pages are hosted on have received a GET request coming from a Google search. Those publishers now know that a particular (cookied?) user could have clicked through to their site. That’s very different from knowing when someone has actually arrived at a particular site.

And that’s why Google host all the AMP pages that they prerender. Given the privacy implications of prerendering non-Google URLs, I must admit that I see their point.

Still, it’s a real shame to miss out on the speed benefit of prerendering:

Prerendering AMP documents leads to substantial improvements in page load times. Page load time can be measured in different ways, but they consistently show that prerendering lets users see the content they want faster. For now, only AMP can provide the privacy preserving prerendering needed for this speed benefit.

A modest proposal

Why is Google’s AMP cache just for AMP pages? (Y’know, apart from the obvious answer that it’s in the name.)

What if Google were allowed to host non-AMP pages? Google search could then prerender those pages just like it currently does for AMP pages. There would be no privacy leaks; everything would happen on the same domain— or or whatever—just as currently happens with AMP pages.

Don’t get me wrong: I’m not suggesting that Google should make a 1:1 model of the web just to prerender search results. I think that the implementation would need to have two important requirements:

  1. Hosting needs to be opt-in.
  2. Only fast pages should be prerendered.

Opting in

Currently, by publishing a page using the AMP format, publishers give implicit approval to Google to host that page on Google’s servers and serve up this Google-hosted version from search results. This has always struck me as being legally iffy. I’ve looked in the AMP documentation to try to find any explicit granting of hosting permission (e.g. “By linking to this JavaScript file, you hereby give Google the right to serve up our copies of your content.”), but no luck. So even with the current situation, I think a clear opt-in for hosting would be beneficial.

This could be a meta element. Maybe something like:

<meta name="caches-allowed" content="google">

This would have the nice benefit of allowing comma-separated values:

<meta name="caches-allowed" content="google, yandex">

(The name is just a strawman, by the way—I’m not suggesting that this is what the final implementation would actually look like.)

If not a meta element, then perhaps this could be part of robots.txt? Although my feeling is that this needs to happen on a document-by-document basis rather than site-wide.

Many people will, quite rightly, never want Google—or anyone else—to host and serve up their content. That’s why it’s so important that this behaviour needs to be opt-in. It’s kind of appalling that the current hosting of AMP pages is opt-in-by-proxy-sort-of.

Criteria for prerendering

Which pages should be blessed with hosting and prerendering? The fast ones. That’s sorta the whole point of AMP. But right now, there’s a lot of resentment by people with already-fast websites who quite rightly feel they shouldn’t have to use the AMP format to benefit from the AMP ecosystem.

Page speed is already a ranking factor. It doesn’t seem like too much of a stretch to extend its benefits to hosting and prerendering. As mentioned above, there are already a few possible metrics to use:

  • Page Speed Index
  • Lighthouse
  • Web Page Test

Ah, but what if a page has good score when it’s indexed, but then gets worse afterwards? Not a problem! The version of the page that’s measured is the same version of the page that gets hosted and prerendered. Google can confidently say “This page is fast!” After all, they’re the ones serving up the page.

That does raise the question of how often Google should check back with the original URL to see if it has changed/worsened/improved. The answer to that question is however long it currently takes to check back in on AMP pages:

Each time a user accesses AMP content from the cache, the content is automatically updated, and the updated version is served to the next user once the content has been cached.


This proposal does not solve the problem with the address bar. You’d still find yourself looking at a page from The Washington Post or The New York Times (or but seeing a completely different URL in your browser. That’s not good, for all the reasons outlined in the AMP letter.

In fact, this proposal could potentially make the situation worse. It would allow even more sites to be impersonated by Google’s URLs. Where currently only AMP pages are bad actors in terms of URL confusion, opening up the AMP cache would allow equal opportunity URL confusion.

What I’m suggesting is definitely not a long-term solution. The long-term solutions currently being investigated are technically tricky and will take quite a while to come to fruition—web packages and signed exchanges. In the meantime, what I’m proposing is a stopgap solution that’s technically a lot simpler. But it won’t solve all the problems with AMP.

This proposal solves one problem—AMP pages being unfairly privileged in search results—but does nothing to solve the other, perhaps more serious problem: the erosion of site identity.


Currently, Google can assess whether a page should be hosted and prerendered by checking to see if it’s a valid AMP page. That test would need to be widened to include a different measurement of performance, but those measurements already exist.

I can see how this assessment might not be as quick as checking for AMP validity. That might affect whether non-AMP pages could be measured quickly enough to end up in the Top Stories carousel, which is, by its nature, time-sensitive. But search results are not necessarily as time-sensitive. Let’s start there.


Currently, AMP pages can be prerendered without fetching anything other than the markup of the AMP page itself. All the CSS is inline. There are no initial requests for other kinds of content like images. That’s because there are no img elements on the page: authors must use amp-img instead. The image itself isn’t loaded until the user is on the page.

If the AMP cache were to be opened up to non-AMP pages, then any content required for prerendering would also need to be hosted on that same domain. Otherwise, there’s privacy leakage.

This definitely introduces an extra level of complexity. Paths to assets within the markup might need to be re-written to point to the Google-hosted equivalents. There would almost certainly need to be a limit on the number of assets allowed. Though, for performance, that’s no bad thing.

Make no mistake, figuring out what to do about assets—style sheets, scripts, and images—is very challenging indeed. Luckily, there are very smart people on the Google AMP team. If that brainpower were to focus on this problem, I am confident they could solve it.


  1. Prerendering of non-Google URLs is problematic for privacy reasons, so Google needs to be able to host pages in order to prerender them.
  2. Currently, that’s only done for pages using the AMP format.
  3. The AMP cache—and with it, prerendering—should be decoupled from the AMP format, and opened up to other fast web pages.

There will be technical challenges, but hopefully nothing insurmountable.

I honestly can’t see what Google have to lose here. If their goal is genuinely to reward fast pages, then opening up their AMP cache to fast non-AMP pages will actively encourage people to make fast web pages (without having to switch over to the AMP format).

I’ve deliberately kept the details vague—what the opt-in should look like; what the speed measurement should be; how to handle assets—I’m sure smarter folks than me can figure that stuff out.

I would really like to know what other people think about this proposal. Obviously, I’d love to hear from members of the Google AMP team. But I’d also love to hear from publishers. And I’d very much like to know what people in the web performance community think about this. (Write a blog post and send me a webmention.)

What am I missing here? What haven’t I thought of? What are the potential pitfalls (and are they any worse than the current acrimonious situation with Google AMP)?

I would really love it if someone with a fast website were in a position to say, “Hey Google, I’m giving you permission to host this page so that it can be prerendered.”

I would really love it if someone with a slow website could say, “Oh, shit! We’d better make our existing website faster or Google won’t host our pages for prerendering.”

And I would dearly love to finally be able to embrace AMP-the-format with a clear conscience. But as long as prerendering is joined at the hip to the AMP format, the injustice of the situation only harms the AMP project.

Google, open up the AMP cache.

Server Timing

Harry wrote a really good article all about the performance measurement Time To First Byte. Time To First Byte: What It Is and Why It Matters:

While a good TTFB doesn’t necessarily mean you will have a fast website, a bad TTFB almost certainly guarantees a slow one.

Time To First Byte has been the chink in my armour over at, especially on the home page. Every time I ran Lighthouse, or some other performance testing tool, I’d get a high score …with some points deducted for taking too long to get that first byte from the server.

Harry’s proposed solution is to set up some Server Timing headers:

With a little bit of extra work spent implementing the Server Timing API, we can begin to measure and surface intricate timings to the front-end, allowing web developers to identify and debug potential bottlenecks previously obscured from view.

I rememberd that Drew wrote an excellent article on Smashing Magazine last year called Measuring Performance With Server Timing:

The job of Server Timing is not to help you actually time activity on your server. You’ll need to do the timing yourself using whatever toolset your backend platform makes available to you. Rather, the purpose of Server Timing is to specify how those measurements can be communicated to the browser.

He even provides some PHP code, which I was able to take wholesale and drop into the codebase for Then I was able to put start/stop points in my code for measuring how long some operations were taking. Then I could output the results of these measurements into Server Timing headers that I could inspect in the “Network” tab of a browser’s dev tools (Chrome is particularly good for displaying Server Timing, so I used that while I was conducting this experiment).

I started with overall database requests. Sure enough, that was where most of the time in time-to-first-byte was being spent.

Then I got more granular. I put start/stop points around specific database calls. By doing this, I was able to zero in on which operations were particularly costly. Once I had done that, I had to figure out how to make the database calls go faster.

Spoiler: I did it by adding an extra index on one particular table. It’s almost always indexes, in my experience, that make the biggest difference to database performance.

I don’t know why it took me so long to get around to messing with Server Timing headers. It has paid off in spades. I wish I had done it sooner.

And now is positively zipping along!