Google no longer requires AMP, but the replacement might be worse
theregister.com
theregister.com
- It's really nice and maybe unprecedented to have an alignment of business, user, eng, and seo goals. Google is using it's monopoly to change roadmap priorities across the industry. The web vitals themselves are unquestionably good for the user.
- Google's guidance on how the scores work and when changes will be incorporated into rankings / carousel eligibility has been overall poor. The info is out there but changes constantly and comes from myriad sources.
- Similarly, the way these scores are measured are a bit of a secret sauce. It's impossible really to model user behavior / devices in a way that ensures you will hit your performance targets. Google provides a lot of tooling/instrumentation, but the measurements sometimes conflict with one another.
I could go on. I'm the mst senior FE guy at this company and it's really consumed most of my work hours since last June, but that's kinda par for the course in publishing. Google swings it's weight around and publishers react.
Which of course you can only see weeks in retrospect.
Or to take it further, it's still nearly impossible to use google's tools and determine that a particular DIV or Ad is causing damaging shift _in aggregate_.
Worse, it's impossible (and probably will always be) to tell the dollar value of getting all the web vitals to "green". They are pretty agressive targets for most publishers I think, but it's unclear exactly how much these fixes are worth (in additional traffic, seo performance, carousel inclusion, etc).
This is a bit "the nature of the beast" when dealing with Google rankings, their most valuable corporate secret, but still, this is a big, hard, expensive job, with relatively clear inputs and measurement, but with totally opaque outputs / ROI.
Fast pages are good, but Google is using their tools like Page Speed to push you to adopt Google’s ideas of fast, like using WebP for images. So now you might have a bunch of cruft in your page to support multiple image types.
Also, as user, I hate lazy loading. I can scroll faster than images load. I’d rather everything load in the beginning, and have images there as I scroll.
That's at least with the HTML5 async image standard in Firefox. It's up to the browser after all, so if yours loads images too slow the issue might be there. And sure, if the images are huge they will take a long time to load, but that's also covered by the core web vitals.
A site I built a while ago is a good test case for that imho, https://www.sustaphones.com/. Long list, many small images, loaded from a CDN. I can maybe provoke a flicker with the scrollbar, but not with regular scrolling.
Edit: That's for broadband. With a 56K modem or something that's of course a different story. But loading a bunch of images in advance on the modern web is also not a good option then.
It depends though on the content page and on how slow the network is exactly, doesn't it? I often enough preferred to have the text already and decide based on that whether to wait for the images. Async with non-jumping base content (which the core web vitals also cover) is great for that.
And with it being a browser feature, even if Google pushes for it the user can always deactivate it. On real browsers at least.
I was critical at first as well. But I tested it on the site I linked and on other sites with bigger images, and it just worked that well that I stand to what I said. This feature is a net-positive and in most scenarios the loading will either not be noticeable, or it will be an advantage overall.
Native lazy loading is broadly supported, it works well, has no overhead, and you can override the behaviour if you always want eager loading.
There's no point in complaining about it that I am able to grasp.
I don't have data on this but I assume that JS lazy loading is dying rapidly because this is not a critical feature that must polyfilled for legacy browsers.
Of course, plenty of sites will keep their JS lazy loading cruft until the next code refactoring / optimization.
But I am very happy that native lazy loading has been standardized and implemented.
Your test case is totally meaningless, the images are under 3kb. Full (non-lazy) image loading is far more important on websites with lots of large images.
You can always construct a theoretical case where it does not work great. But on those pages and connections loading all the images in advance will usually also lead to a bad experience.
The typical WordPress lazy loading script doesn't start loading the image until it is scrolling into view. Try a typical food blog, scroll moderately slowly and you'll still get to see every single image load.
edit: your example works great, but is in outlier in the implementation, and benefits from tiny images that load quickly.
My background that made me write the comment above is that I implemented lazy loading for a feedreader in JS, with predictive early loading to exactly avoid this problem, and I did test the browser feature just recently again. In those scenarios lazy loading really does work well enough for my statement to be true, according to my testing. For common connection speeds and image sizes only of course. Though I'd argue it also helps in less optimal scenarios exactly because it somewhat decouples the image loading from the page load - that that overall experience was horrible on long threads with many images in the feedreader was why the lazy loading feature was added, the contrast was stark.
I should have defined that better in the comment above and given your experience more validity. Next time :)
This is an example of what I am talking about.
https://clayimports.com/collections/colors
It "passes" the google lazy load audit, but is a terrible user experience. Doing no image load optimization would be much better than what they are doing presumably to please google.
He preferred to wait all the time until images are loaded to start reading articles, with a flaky slow connection, or to try start reading while parts of the site still load? Especially on a slow connection not having to load the images immediately, or at all if the visitor does not scroll down as much, can make browsing so much better.
The main problem the customer I was referring to had is that the user interfaces for these websites would not function until they finished loading after 30 to 90 seconds, despite the fact that the user interfaces rarely changed. These were for cloud based accounting and scheduling platforms the business was required to use.
Developers really should have to try to use the systems they build on systems that have low bandwidth / high latency network connections rather than the typical fibre fed office environment. If you can build a website that works well for that use case, it's probably going to be incredibly snappy on a fast connection. Judging by how bloated the web is these days, that is rarely the case in today's world.
My definition of broadband is correct, and 1.8Mbps does not match it. Where I live 1.8Mbps just is not legal broadband. In the US that also does not count as broadband anymore, see https://broadbandnow.com/report/fcc-broadband-definition/, and it hasn't for over a decade.
I must say that the FCC really messed up by redefining broadband when they were the ones that had originally defined it in terms of RF spectrum usage. Here in Canada, the telecom regulatory body set the Basic Service Objective to be 50 Mbps download and 10 Mbps upload. They didn't try to redefine what broadband is and overload a term that already had meaning in the telecommunications space.
Personally I think latency is more important with modern internet connections. 10 Mbps is still quite usable for a lot of things, but not if it's 500ms RTT via a satellite. Sadly most telecom regulators ignore latency.
its look at any site in chrome dev tools.
Lazy-loading can be implemented sloppily, but browser-native lazy loading is now good enough to be considered optimal.
The problem is that Google Page Speed dings you for not having lazy loading, and people add crappy JavaScript lazy loading that satisfies Google, but makes the user experience worse.
WebP is good iff it improves rendering times (which it does). That's not google's fault really.
The problem has been, until recently, Safari didn't support WebP, so you had to implement a fallback to jpg. More complicated HTML, more images to compress and manage, all for a very marginal improvement.
I assume that those sites will vanish quickly though.
"Lazy loading" interactive content is an entirely different thing.
Shopping pages with tons of product images, Infinite scroll (yuck) are different...
Google ; New Breadcrumbs issue detected for site xxxxx 24 Jun 2021•Breadcrumbs
So I look into it, knowing I'd fixed that ages ago....
Google ; Last crawled : 13 Nov 2019
Triggering stuff because you eventually got round to looking at it 19 months later? and it's all like that full of massive holes that make zero sense.
Google; "This product doesn't show how many reviews it has"... Me; "It has no reviews" Google; "ERROR! ERROR! ERROR! ERROR! zero is not a valid number!!!!" Me; "Fuck you Google" Google; "This product doesn't show how many reviews it has"...
I'm not even going to get into how shitty pagespeed insights is* - it's not even close to lighthouse. It tells you to increase your caching but doesn't use caching in performance measurements. I presume google has a giant ass that it pulls most of the figures out of on each run. I live in an area of the UK with one of the worst mobile phone signals known to man, pages load in < 2 seconds google consistently claims at least double.
I really don't enjoy spending hours of my days having to check that Google haven't moved the goalposts yet again.
* I lied, I am.
Having "page performance" as a non-trivial weighted factor in how pages are ranked just makes sense to me.
Even if CWV is somewhat flawed, I'd much rather optimize for that than give control of my pages to Google ala AMP.
[1] https://support.google.com/webmasters/answer/9205520?hl=en
This means that your hand-coded-in-vim static pages will rank near the top.
I use a somewhat hand-coded-in-vim approach, but there's some post-processing that adds width/height attributes to <img> tags which mitigates this particular problem.
Edit: Not sure why everyone is telling me to add "height" and "width" attributes, because that's what I wrote in the comment, above. Just to explain how this works--the width and height attributes describe the "intrinsic size" of the image.
Those are used to determine the aspect ratio, so even when resized via CSS the browser knows how much space to reserve in the layout.
The aspect ratio is determined by a combination of CSS rules and the intrinsic size.
The CSS sizing will still take precedence but the height/width attributes let the browser know the aspect ratio without having to download the image so it can layout the page faster.
It's surprisingly hard to set up good images in a hand-coded-in-vim website. You're supposed to include multiple image sizes and set up the code for loading each in HTML. So what could have been simply adding an <img> tag now turns into a chore involving ImageMagick.
I know this is only orthogonal to AMP, but it sounds like the kind of thing Core Web Vitals was meant to disincentivize.
I've also noticed that particular bit of irony. In fact it's the longest-running CLS issue I've seen anywhere.
Just another veiled article criticizing Google's dominance rather than a technical one.
There has to be a way of prioritizing some of these sites over others. Same goes for sports, celebrity/gossip, news, etc. That would be most of the open web right there.
Google also moved to mobile-first indexing some years ago. Was that unfair to mom-and-pop shops that didn't have a responsively designed site? Perhaps, but if the website is something that brings you business, it's a tool that should be sharpened when dull, and replaced when its rusty.
Those that display fewer ads and trackers? Many people would like that but somebody would really not.
The overwhelming majority of websites being created today include some kind of tracker, even if it's a single analytics script. And knowing the HN audience, that's still one tracker too many.
I have a gigabit connection, so it is not a question about not downloading fast enough, I can browse wikipedia this way. It is a matter of not using so much javascript.
Core Web Vitals, as a publisher myself, are a nightmare. But at least we know what goes wrong as it's clearly laid right in front of the developer's eyes. So we go and fix thing after thing. And some pages just won't get fixed because of some ads.
The point is CWV do make the web better. Those vital metrics are actually making sense (CLS, LCP).
I understand the reticence and reserve at who is behind both these efforts that AMP and CWV are, but they forced me to make my websites faster, cleaner.
Was it a panicky few months trying to get this sorted? Yes. Did we have to speak to ad networks so they fix up some ad delivery? Yes. Will Google really rank us better thanks to that? Maybe. Couldn't we have done it from our own accord? The incentive wasn't big enough I guess.
But today, the fact is, the experience on all of our websites is very much clearly improved.
How? No user has ever chosen to open a link as AMP.
> so I won't be the one recommending that.
This makes me wonder if the choice between leaving in ads that lower page rank due to CWV and removing some ads to get a better page rank will lead to less advertising overall. I can only hope, I guess.
Waiting once is fine; waiting two times means I can't do something else while it's loading
Web developers don't suck at making fast websites. Publishers demand tons of ads and tracking scripts. If the person signing the paychecks wants the page to have 15MB of sketchy third-party JS, then the page will have 15MB of sketchy third-party JS. AMP succeeded because Google brandished a stick that publishers cared about more: exclusion from the Top News carousel.
I'd also note that the whole reason this ad tech explosion even happened is a race to the bottom that Google itself facilitated. Google is selling us solutions to a problem that Google had a huge hand in causing.
I was asked by management to add Google Tag Manager to the company site. When I saw that GTM included 15-20 tracking scripts I told them that unfortunately we can't include it until we have ensured that all of them act in accordance with GDPR. Legal was involved and agreed. Marketing gave up.
You're technically correct, but people who write pay checks often can be argued with. They often don't want to be on record taking decisions which can cause issues, and legal typically don't want to sign of on including 15MB unknown scripts on your site accepting credit cards.
That's true, but the proliferation of ad tech junk is strong evidence that this is the exception rather than the rule. Capital will persuade labor to do its bidding far more often than not.
Or are we going to talk about how online advertising has been the economic backbone of the web? Not that I'm advocating for it, by hiding the true cost of things, publishers have brought this onto themselves really.
So no more ads? That's probably going to kill a lot of sites, whether they are contributing anything to mankind or not.
Or we can acknowledge that advertisment is there to stay and we instead develop a proper framework to make it work sanely rather than all the duct-taping we've done so far.
The real failure is the unwillingness to confront the fact that ads were there to stay and should have been baked into a W3C standard ages ago instead but here we are.
Either way, if their process of reaching their end goal included steps which legal dismissed then they needed to go back to the drawing board and figure out a new strategy, right? It would be strange for me who doesn't work in their profession come and tell them how to solve their issues.
You're not telling them, you're helping them.
I guess I can't get past the idea of telling someone I work with that their solution won't work and then being like, "sorry you're on your own".
Although I'm sure you don't care, the disdain you show for marketing wouldn't make me super excited to work with you.
Like @djtriptych I do a lot of work on a major brand sites its poor FE design and implementation.
You can at least push back on the product somewhat, or use something like Segment to fan-out. But once Marketing gets GTM in place, all bets are off - they're doing god knows what to the site/application, completely outside of the purview of the usual engineering process, because they can inject whatever they hell they want into the page.
GTM is essentially a backdoor but it gets a pass because marketing.
Web devs emphatically, and by definition, do not suck at developing websites.
Sorry but I don’t think that’s “worse” than AMP at all. Google was in charge of that and it hijacked URLs.
Clickbait.
Why not just use overall page size and ttfb instead of this nonsensical metric that isn't a great measurement, in my experience, of website performance/user experience; outside of keeping everyone guessing what Google might do next, what's the point? I also find it extremely ironic that many of Google's own products are complete failures on the metric.
I've even found that you can trick the metric by essentially loading a blank page where a browser will feel the page is fully loaded, then having a delay that loads the page.
But there's a metric specifically for measuring load times. https://web.dev/lcp/
Did no one in Ads team get the memo to optimize performance?
They were too busy raking in the billions.
So an even more disproportionate amount of site performance work will now be focused just on what makes Chrome happy. Experimental Chrome features that boost performance metrics will be embraced in the name of SEO. Safari and Firefox can't offer that.
It's another way Google uses their control of the search market to cement their browser as the de facto standard.
For example, don't hire people from Region A because they can't access the best resources located in Region B.
This the future foreign internet companies chose when they put profits above all else.
Imagine that I call you over the phone, old school style, with voice. Would you or would you not want the governments on either end of the phone line to eavesdrop on our conversation so they can make sure it meets their local laws? What if we communicated by SMS? Email? Posting on a message board?
Some people will have an "obviously yes, I want the government to be in charge, are you dumb?" reaction. Others will say "of course I want privacy, are you dumb?"
Early Internet chose privacy. Modern internet is increasingly choosing regulation. But this is far from the obvious truth you are claiming it is, if only because the prevailing opinion changed in recent memory.
If only as a way of stopping foreign criminal gangs from scamming the vulnerable.
The main reason this can't happen is because of international agreements that were made long before the internet age or this type of crime became a thing.
It's amusing how on another thread the general consensus may be a celebration of Let's Encrypt while over here I seem to come off as crazy (and thus deserving of downvotes simply for saying "not everyone agrees with you").
Do you want the government to bug you and watch every interaction you have in your local store to find out if the store is breaking the law? But then how does the government know if the store is just not paying their taxes, or selling drugs?
Laws can be enforced without resorting to wiretapping everyone by just not accepting you can't catch every single violation; that true both online and offline.
Whether or not you think it’s correct, the obvious and inevitable conclusion is that the web will be siloed off and cease to be a global network.
In any case, once again it’s irrelevant. Whatever your justification for it, this behavior will lead to internet siloization.
Plenty of countries require you to register for sales tax once you establish a presence in their country, whether you sell goods or services. No business operating in this environment thinks it’s unreasonable to do so. Once again “their house, their rules.”
No, you are confusing sending information with “doing business”.
Why is it unjust when literally the entire U.K. population is in support of this position?
Obviously not "literally the entire U.K. population" if someone in the U.K. isn't following the rule.
This is a complex subject and I'm not going to get into it here, but the Cliff's Notes version is basically that the U.K. government is not a party to this transaction, is not harmed by it, and does not represent (as in: having a formal, revocable agent/principal relationship with) anyone who is either a party to the transaction or harmed by it, and thus has no standing to interfere. The justice or injustice of the matter is unaffected by whether the government's interference would be popular.
Longer version: governments adopt rules based on what works for them and their population. This includes rules that govern what can and can’t be sold in a business transaction. This is a principle as old as government itself. It makes sense that governments then apply these rules to things going in and out of its borders. It would be nuts to ban the sale of guns inside a country but allow them to be sold into the country, for example.
If we're talking about a "business transaction" as in an exchange of physical goods across the border, then I agree. I said that it was unjust, not that it was without precedent. You seem to be under the mistaken impression that I believe in the concept of government itself. The core of what government is, and does, is unjust. Arguing that governments have always done things this way carries zero weight with me.
Applying these rules (or any rules) to the non-commercial exchange of information, or even to commercial services involving no exchange of physical property, is the recent, and more immediately concerning, development.
> It would be nuts to ban the sale of guns inside a country…
I agree with you up to that point. But if they want to ban the import of "wireless handheld hold punchers" or any other contraband they should do that at the border, by stopping the shipments—which at that point consist of the buyer inside the country attempting to import their own property after the sale—and not by attempting to impose their internal rules on foreign sellers. Residents could buy them but would have to keep them outside the country, perhaps using them only while visiting the country the goods were in at the time of the sale, or somewhere else where they are legal.
I’m not interested in whether you believe in government or not because that’s debating a fantasy world that simply doesn’t exist.
You're not in their "house". You're just communicating with someone who is. If some foreign government has a problem with that communication they should take it up with the party who is physically within their jurisdiction—not that I believe they have any just standing there either.
Why does this magically change when someone utters the words “the internet?”
You are not doing business in another country. First, there may not be any "business" involved at all. More importantly, however, you are not in the other country. The other person is. If physical property is involved then someone—not necessarily you—is going to need to worry about import/export regulations when moving it across the border. When it comes to virtual services and websites, however, those concerns rooted in physical transportation do not apply.
> Why does this magically change when someone utters the words “the internet?”
It makes no difference whether you're communicating over the Internet or by phone or two-way radio or postal mail or carrier pigeon, or by shouting to each other across the border. So long as you're only exchanging data and not physical goods they can implement technical measures to block the communication at the border, or order their own residents not to communicate with you—in which case I would advise them to make arrangements leave the country posthaste—but you yourself remain outside their jurisdiction.
This is something that’s already heavily regulated for other industries e.g. financial services. There’s nothing novel about the GDPR in that aspect.
Services which involve nothing but data, and thus still fall under the category of "data moving over borders". And, consequently, freedom of speech.
May I introduce you to services such as Wikipedia.com and news.ycombinator.com?
Which is literally just data.
> a forum through you can interact with other people
Which is just a means of exchanging data.
What is missing in both cases is any transportation of people or material across the border.
Regardless of whether or not your model actually makes sense, an inevitable consequence of your model is that the internet will become siloed at a national level. It sounds like you’re fine with that though.
Honouring DNT should be trivial, no user interaction needed to stop tracking and therefore the need for 99% of these popups.
Many websites prefer to destroy their usability to get your consent in various deceptive ways. You can definitely have technical cookies and no cookies banner.
And if they did, what do you think the public reaction would be to such data sharing?
The real problem here is not 'gamed search engines' but too much centralisation, with Google owning the search engine, the browser, and the ad platform. This solution makes that problem worse, not better.
It's like I've fallen into a crazy parallel dimension. For more than *five years* HN commenters have been ranting about how AMP is not needed, and all that's needed is using page speed as the signal. And now when that happens, it's suddenly an outrage.
Google has such a dominant position in both search and browsers that, yeah, they're going to need to work extra hard to make sure one doesn't unfairly advantage the other.
The morale of the story is it's impossible to make everyone happy.
But there's no contradiction in asking for a neutral way to include performance in the rankings and then being disappointed when the solution considers only performance of Google's own browser.
Shifting from one technology to another is not a solution when all the options vacuum and hoard data to further entrench Google's position. A real choice would be Google splitting the different parts of their business model they are leveraging to lock out competitors, but since that's not good for their shareholders, I don't expect to see it any time soon.
But yes, I think it's plausible that other browser makers could allow users to opt in to this sort of data reporting. It would lead to a more equitable web.
I'm also not totally convinced that page performance should play a significant role in rankings in the first place.
Second, you wouldn't complain that it's an unfair advantage to pages in the Google AMP cache, due to a better network location? An unfair advantage to pages hosted on GCP? An unfair disadvantage to iOS due to there being no iOS server hardware that could be used for testing?
Easy. Google should not be operating the monopoly search engine _and_ the monopoly web browser. They need to spin off search or Chrome as a separate company. Then they can source their data from wherever they like. Then every browser is an "other browser".
I worry that if core vitals becomes too strong of a signal it will over homogenize things to the point that we’re further ostracizing parts of the web that aren’t built like fast results pages to get an answer and bounce off of. We need more differing alternatives to Google.
Just change the algorithm to reward fast websites even more.
If page speed currently counts for say 5% of a site's value in the algorithm, bump it up to 10%, 15%, 25% etc, whatever it takes until sites that pump in piles of garbage JS trackers and monstrosities start removing that stuff.
Sites that insist on leaving in garbage trackers and bloated designs will not show up as much. Problem solved without dictating technology.
> Unfortunately, there are problems with AMP's replacement as well. And those problems go right back to what was wrong with AMP in the first place: Google is in charge of it.
So the problem isn't core web vitals nor AMP itself, but the fact that everyone uses Google and they control the results on google.com.
We can now go back to our regularly scheduled discussion about how this is bad.
Some people are so deeply lost in their hatred of AMP that they cannot even consider for a moment that maybe, just maybe, AMP actually does work well for some people.
I'm glad it's no longer required, and if people are correct that AMP truly sucks, then surely it'll slowly fade away over time.
Of course you can over engineer anything, but why would you when [almost] everything you need is easy to grab a la carte.
- no JavaScript
- you get 100kB for HTML and CSS
- all decorative and non-essential images are included in those 100kB
Happy hacking!
Also, I'm not sure this is well known, but if you navigate to a web page from the carousel, Google traps the left and right swipe events on YOUR page. Either sends to user to a competitor. That's a pretty big compromise you would only make because the traffic boost from the carousel was worth it.
https://arstechnica.com/gadgets/2021/06/qualcomms-new-snapdr...
Edit: As for Conde Nast, I found this article: https://technology.condenast.com/story/the-why-and-how-of-go... The diagram[1] suggests the Google AMP cache is (or was) in front of users, and they also say this:
"AMP helps us satisfy these needs. AMP increases the visibility and discoverability of our content by allowing it to be included in Google’s Top News Carousel, as well as improving the experience of regular Google search results"
That sounds a lot like what I said.
[1] https://media.condenast.io/photos/59a57e16cf742825d0b02891/m...
I click it, and get this: https://imgur.com/a/2W3oWSW
The top bar is Google controlled. The url in the omnibar is a google url. So I'm not seeing how Conde Nast is a good example here. Before they very recently rescinded the requirement, using a non-Google AMP cache meant you didn't get into the carousel.
Edited to add: An example of where the same content is not framed by the same search engine is on Chrome Mobile's new tab page "articles for you". Many of these are directly links to unframed AMP content, such as what I'm seeing right now: seekingalpha.com/amp/..., thehill.com/...?amp etc
For Chrome, yes.
My assumption is that AMP is most heavily used by news organizations because of the requirement (recently rescinded) for carousel placement. Anecdotal, but I've talked with people at these organizations, and they definitely don't like AMP. They tolerate AMP solely for the carousel placement. That requirement is gone now. But, given the low margins in those places, it may be some time before they migrate off.
You must be a mind reader, then, because nowhere in the article does the author state or imply that. Indeed, they go out of their way to point out why it's unlikely in the extreme that any publisher will bother, and points out why that's essentially a victory for Google -- indeed this might be the essential point that the author is making in this article.
It feels like we read different articles, because the in one I read the author spent a lot of time explaining that AMP would probably be around for a while for a number of predictable reasons.
The idea that PSI preloads the page before starting the test doesn't make much sense at all, and doesn't mesh with my own experiences (in the same space).
*In reality, about 1% of human knowledge, by some estimates.
https://developers.google.com/search/blog/2020/11/timing-for...
http://www.earth.org.uk/note-on-site-technicals-49.html#2021...
HTML is not bad and it is easier to parse than markdown, the trouble is javascript and the ways ads abuse them. Strip that out, set the size of images and you can create a better experience for your users without changing all browsers.
However if we can get wild requests for new content types, then I would love text/latex, so that I could create a website that fits the real size of your browser and render a gorgeous output.
In the last decade, I’ve seen browsers drop content types (FTP) and adopt media formats (WebP). But no new publication formats. If my browser can render PDF, why shouldn’t it render Markdown?
The power of the web lies in its decentralisation,
it lies with its messiness, it lies with its edge
nodes – that is, with you and me.