It's a product of many cooks and their brilliant ideas and KPIs, a social network for devs and code being the most "brilliant" of them all. For day to day dev operations is something so mediocre even Gitlab looks like the golden standard compared to Github.
And no, the problem is not "Rails" or [ insert any other tech BS to deflect the real problems ].
For my sins I occasionally create large PRs (> 1,000 files) in GitHub, and teammates (who mostly all use Chrome) will sometimes say "I'll approve once it loads for me..."
Unrealistic timelines, implementing what should be backend logic in frontend, there's a bunch of ways SPA's tend to be a trap. Was react a bad idea? Can anyone point to a single well made react app?
It's hard to know which member of the duopoly is more guilty for breaking GitHub for me, but I find that blaming both often guarantees success.
I could like, buy a new computer and stuff. But you know, the whole Turing complete thing feels like a lie in the age of planned obsolescence. So web standards are too.
Does anyone have concrete information?
I had to alter basically every aspect of how I interact with it because of how fucking slow it is! I still can't shake the sense that it's about to go down or that I've done something wrong every time I click something and nothing happens for several seconds.
At the very least, I wish they set it to auto.
I have an ever growing directory listing using SolidJS, and it's up to about 25,000 items. Safari macOS and iOS two major versions ago actually handled it well. After the last major update, my phone rendered it faster than an m1 MacBook Pro.
On random site, Navigate to GitHub repo, navigate to file in repo, and hit back, and I'm on the random site, hit forward and I'm on the file.
So annoying.
One of a large handful of issues I've encountered post react conversion
Good to know others are feeling it too, hopefully it can get resolved soon. In the mean time, i'll try my PR reviews on FF.
Update: Just tested my big PR (+8,661, -1,657) on FF and it worked like a charm!
You really can't escape the enshittification.
The problem is they abandoned rails for react. The old SSR GitHub experience was very good. You could review massive PRs on any machine before they made the move.
Too bad Phabricator is maintenance-only now https://en.m.wikipedia.org/wiki/Phabricator
Clean code argues that instead of total rewrites you should focus on gradual improvements over time, refactor code so that overtime you pay off the dividends, without re-living through all the bugs you lived through 5 years ago that you don't recall the resolution of. Every rewrite project I've ever worked on, we run into bugs we had already fixed years prior, or the team before me has.
There are times when a total rewrite might be the best and only options such as deprecated platforms (think of like Visual Basic 6 apps that will never get threading).
What frustrates me more is that GitHub used to be open to browse, and the search worked, now in their effort to force you to make an account (I HAVE LIKE TEN ALREADY) and force you to login, they include a few "dark patterns" where parts of search don't work at all.
I see loading spanner everywhere and even the page transition take ages compared to before.
I am not sure what metric they are using justify ditching the perfectly working SSR they used before.
Slow as hell and the Safari search function stopped working. I loaded the same url on Firefox and it was insta-fast.
Then some charlatan thought to embrace the React hype and it became terrible to say the least.
The solution is a test that fails when Chrome and Safari have substantially different render times.
My CPU goes to 100% and fans roaring every time I load the dashboard and transactions. I can barely click on customers/subscriptions/etc. I can't be the only one...
Don't listen to the opinions of the developers writing this code. Listen to the opinions of the people making these tech stack decisions.
Everything else is a distant second, which is why you get shitty performance, developers who cannot measure things. It also explains why when you ask the developers about any of this you get bizarre cognitive complexity for answers. The developers, in most cases, know what they need to do to be hired and cannot work outside those lanes and yet simultaneously have an awareness of various limitations of what they release. They know the result is slow, likely has accessibility problems, and scales poorly, and so on but their primary concern is retaining employment.
- Project managers putting constant pressure on developers to deliver as fast as possible. It doesn't even matter if velocity will be lost in the future, or if the company might lose customers, or even if it breaks the law.
- Developers pushing back on things that can backfire and burning political capital and causing constant burnout. And when things DO backfire, the developer is to blame for letting it happen and not having pushed it more in the first place.
- Developers who learned that the only way to win is by not giving a single fuck, and just trucking on through the tasks without much thought.
This might sound highly cynical, but unfortunately this is what it has become.
Developers are way too isolated from the end result, and accountability is non-existent for PMs who isolate devs from the result, because "isolating developers" is seem as their only job.
EDIT: This is a cultural problem that can't be solved by individual contributors or by middle management without raising hell and putting a target on their backs. Only cultural change enforced by C-Levels is able to change this, but this is *not* in the interest of most CEOs or CTOs.
Todays version is: "You will get fired unless you use React".
So every site now uses React no matter if the end result is a dog slow Github.
Bad developers looks at "what are everybody else using?".
Good developers looks at "what is the best and simplest (KISS) tool for this?"
If you put a lot of momentum behind a product with that mentality you get features piled on tech debt, no one gets enthusiastic about paying that down because it was done by some prior team you have no understanding of and it gets in the way of what management wants, which is more features so they can get bonuses.
Speaking up about it gets you shouted down and thrown on a performance improvement plan because you aren't aligned with your capitalist masters.
The short answer is: no, they don't. Google Cloud relied upon some Googlers happening to be Firefox users. We definitely didn't have a "machine farm" of computers running relevant OS and browser versions to test the UI against (that exists in Google for some teams and some projects, but it's not an "every project must have one" kind of resource). When a major performance regression was introduced (in Firefox only) in a UI my team was responsible for once, we had a ticket filed that was about as low-priority as you can file a ticket. The solution? Mozilla patched their rendering engine two minor versions later and the problem went away.
I put more than zero effort into fixing it, but tl;dr I had to chase the problem all the way to debugging the browser rendering engine itself via a build-from-source, and since nobody had set one of those up for the team and it was the first time I was doing it myself, I didn't get very far; Google's own in-house security got in the way of installing the relevant components to make it happen, I had to understand how to build Firefox from source in the first place, my personal machine was slow for the task (most of Google's builds are farm-based; compilation happens on servers and is cached, not on local machines).
I simply ran out of time; Mozilla fixed the issue before I could. And, absolutely, I don't expect it would have been promotion-notable that I'd pursued the issue (especially since the solution of "procrastinating until the other company fixes it" would have cost the company 0 eng-hours).
I can't speak for GitHub / Microsoft, but Google nominally supports the N (I think N=2) most recent browser versions for Safari, Edge, Chrome, Firefox, but "supports" can, indeed, mean "if Firefox pushes a change that breaks our UI... Well, you've got three other browsers you could use instead. At least." And of course, yes, issues with Chrome performance end up high priority because they interfere with the average in-house developer experience.
There are of course performant react apps out there. What Steve did with tldraw is amazing.
However, the vast majority of the apps out there are garbage since the framework itself is terribly inefficient.
I still love it! Works great, makes sense, is fast...
Good ol’ SSR - but eventually users and PMs start requesting features that can only be implemented with an SPA system, and I (begrudgingly) accept their arguments.
In my role (of many) as technical architect for my org, and as an act of resistance (and possibly to intentionally sabotage LLMs taking over), I opted for hybrid SSR + Svelte - it’s working well for us.
Which, it seems, was a result of the M$ acquisition: https://muan.co/posts/javascript
Their "solution" was to enable SSR for us ranters' accounts.
My memory is fuzzy but I think it was on phab that I discovered and loved to use stacked merges. This is where you have a merge request into another open merge request etc. Super useful. Miss that in the git world.
Gitlab is anything but light, by default tends to be slow, but surprisingly fast with a good server ( nothing crazy, but big ) and caching.
If a developer has to put up a fight in order to push back against the irresponsibility of a non-technical person, they by definition don't have ownership.
Is it your theory that working on large projects was better when you had communist masters? That seems inconsistent with everything we know, e.g. quotas enforced my mass murder.
My guess is that it's more about organizations (your first paragraph) and less about capitalism (your last paragraph).
On react, it's funny that sites where the frontend part is really crucial tend to move away from generic frameworks and do really custom stuff to optimize. I'm thinking about Notion, or Google Sheets, or Figma, where the web interface is everything and pretty early on they just bypass the frontend stacks generally used by the industry.
The problem isn't React. The problem are KPIs and unrealistic timeline. It is the same then ever. Not a fault of React at all.
Planned obsolescence is some of it, some of it is abstractions making it easier for more people to make software (at the cost of using significantly more compute) and Moore’s law being able to support those abstraction layers. Just imagine if every piece of software had to be written in C, the world would look a whole lot different.
I also think we’ve gone a bit too far into abstraction land, but hey, that’s where we are and it’s unlikely we are going back.
Turing completeness is almost an unrelated concept in all of this if you ask me, and if anything it’s because of completeness that has driven higher and higher memory and compute requirements.
It's just easier to blame the tools (or companies!) you already hate.
A very simple example: migrating from JavaEE to JakartaEE. Every single Java source file has to have the imports changed from "javax." to "jakarta.", which can easily be thousands of files. It's also easy to review (and any file which missed that change will fail when compiling on the CI).
I don’t know if that’s a good or realistic rule for most projects, but I imagine for performant types of applications, that’s exactly what it takes to prevent eventual slowdown.
The Cloud to make single-digit-seconds operations on a local Raspberry Pi 2 and home Internet take a few minutes.
What a time to be alive.
GitHub issues was so simple and now they keep shoving features into it.
Why has no one learned to not become Jira? You gotta say no sometimes.
That test will be disabled for being flaky in under a week because the CI runners have contention with other jobs, causing them to randomly be slower and flake, and the frontend team does not want to waste time investigating flakes.
"Just have dedicated runners with guaranteed CPU performance", but that's the CI platform team's issue, the frontend and testing teams can't fix it, and the CI infra team won't prioritize it for a minimum of 5 years.
React can have all the niceties and optimization in the world, but that fails when its users insist on using it incorrectly, building huge tangled messy components and then wondering why a click takes 1.3 seconds to deliver feedback.
Back in the day (I was a junior dev) this was easier than grappling with React hooks today:
BOOL CMainDialog::OnInitDialog()
{
CDialogEx::OnInitDialog();
m_pPropertySheet = new CMyPropertySheet(_T("My Tabbed Dialog"), this);
m_pPropertySheet->Create(this, WS_CHILD | WS_VISIBLE, WS_EX_CONTROLPARENT);
CRect rectMainDialog;
GetClientRect(&rectMainDialog);
CRect rectPropertySheet(10, 10, rectMainDialog.Width() - 20, rectMainDialog.Height() - 20);
m_pPropertySheet->MoveWindow(rectPropertySheet);
return TRUE;
}I don’t think the culprit apps would have substantially better UX if they were rendered on the server, because these issues tend to be a consequence of devs being pressured to rapidly release new features without regard to quality.
In case you're one of today's lucky 10,000, OpenCore Legacy Patcher supports Macs going to back as far as 2007: https://github.com/dortania/OpenCore-Legacy-Patcher
I know some people feel like Apple is aggressive in this respect, but that's an 8 year old version of a browser. That's like taking off all of the locks on your house, leaving the doors and windows open all while expecting your house to never have uninvited guests.
[1]: https://yoyo-code.com/why-is-github-ui-getting-so-much-slowe...
But CSS has bit me with heavy pages (causing a few seconds of lag that even devtools debugging/logging didn't point towards). We know wildcard selectors can impact performance, but in my case there were many open ended selectors like `:not(.what) .ever` where the `:not()` not being attached to anything made it act like a wildcard with conditions. Using `:has()` will do the same with additional overhead. Safari was the worst at handling large pages and these types of selectors and I noticed more sluggishness 2-3 years ago.
I actually have been trying to figure out how to get my React application (unreleased) to perform less laggy in Safari than it does in Firefox/Chrome, and it seems like it is related to all the damn DOM elements. This sucks. Virtualizing viewports adds loads of complexity and breaks some built-in browser features, so I generally prefer not to do it. But, at least in my case, Safari seems to struggle with doing certain layout operations with a shit load of elements more than Chrome and Firefox do.
"Rename 'CustomerEmailAddress' to 'CustomerEmail'"
"Upgrade 3rd party API from v3 to v4"
I genuinely don't get this notion of a "max # of files in a PR". It all comes off to me as post hoc justification of really shitty technology decisions at GitHub.
But there is also the Safari Technology Preview, which installs as a separate app, but is also a bit more unstable. Similar to Chrome Canary.
it's Microsoft, so the answer is: buy a new computer
(which comes with a bundled Windows license)
If you actually load up a ~2015 version of Jira on today’s hardware it’s basically instant.
A lot of the time we just break the branch permissions on the repo we are using and run release branches without PRs and ignore the entire web interface.
Old GitHub was very light on features, whereas the new UIs are way more curated on the surface.
Unfortunately all of this brings in tons of complexity. It doesn't help that there are a lot of junior developers working on it, clearly.
The fact that they have this ability / awareness and haven't completely reverted by now is shocking to me.
I’m sure you could make something work better as a SPA, but nobody does.
Github's code view page has been unreasonably slow for the last several years ever since they migrated away from Rails for no apparent reason.
In this very thread there's some asshole using the word "memoization" when "caching" would have been fine.
https://chromewebstore.google.com/detail/make-github-great-a...
Normally, you be able to debug selector matching performance (and in general, see how much style computation costs you), so it's a bit weird if you have phantom multi-second delays.
By all means. It sometimes feels like React is more the symptom than the actual issue, though.
Personally I generally just like having less code; generally makes for fewer footguns. But that's an incredibly hard sell in general (and of course not the entire story).
I would rather just see the steps you ran to generate the diff and review that instead.
if they were forced to use slow machines, they would not be able to put out crap like that
I assume this is fallout from dealing with LLM content scrapers.
Available data confirms that SPA tends to perform worse than classic SSR.
Gitea is an example I like because it stores the repository as a bare repository, the same as if I did git clone --bare. I bring it up because when I stopped running Gitea, I could easily go in to the data and backup all the repositories an easily reuse them somewhere else.
But I guess the problem is that every single development position has been converging into this.
The only times in my career as a developer where I was 100% happy was when there was no career PM. Sales, customers, end-users, an engineering manager, another manager, a business owner, a random employee, some rando right out the street... All of those were way better product owners than career PMs in my 25 years of experience.
This is not exactly about competence of the category, it's just about what fits and doesn't. Software development ONLY work when there is a balance of power. PMs have leverage that developers rarely have.
I come from Electrical Engineering. Engineering requires responsibility, but responsibility requires the ability to say "no". PMs, when part of a multi-disciplinary team, make this borderline impossible, and make "being an engineer" synonymous with putting a target on your back.
For instance, the GP could be a proponent of self-management, and the statement would be coherent (an indictment of leaders within capitalism) without supposing anything about communism.
Svelte is ok. It could have been great but the api for their version of observables is a disaster (which I hope they eventually fix). Sveltekit is half baked and convoluted and I strongly advise not touching it.
IMO it's the MAIN thing to understand about React—how it renders.
Regardless, now I'm the one with egg on my face since the new compiler promises to eventually remove the need for manual memoization almost entirely. The "almost" still fills me with fear
So GitHub is usable but there are a number of UI layout issues and searching within a file is sometimes a mess (eg, highlighting the wrong text, rendering text incorrectly, etc. maybe that's true for all browsers. you're better off viewing a file as text in raw mode)
As an aside, I was an employee around then and I vividly remember that the next half there was a topline goal to improve web speed. Hmmmm, I wonder what could have happened?
Depending on where you live (or what websites you visit) it's not unreasonable.
I haven't been able to load it yet to actually check out these hip new features, it just crashes my browser, but I'm sure they must be great?
And this is something browsers don't treat as bugs. You can crash any browser's tab by just exhausting its allocated memory
It was being hosted on another continent. It was written in PHP. It was rendering server-side with just some light JS on my end.
That used to be the norm.
> publicly disseminate information regarding the performance of the Cloud Products
https://web.archive.org/web/20210624221204/https://www.atlas...
network.http.referer.XOriginPolicy = 1
https://we.phorge.it/phame/post/view/8/anonymous_cloning_dis... https://we.phorge.it/phame/post/view/9/anonymous_cloning_has...
At any rate your point doesn't make any sense. The same point indicts all leaders, it has nothing to do with capitalism. It's like saying something indicts a specific race of people when it applies to all people equally.
GitLab: https://docs.gitlab.com/administration/gitaly/praefect/
GitHub: https://github.blog/engineering/infrastructure/stretching-sp...
Firefox doesn't work on Windows 7 anymore but installing Firefox is still a hell of a lot better than sticking to IE.
The usual response is something like "if you're correct, wouldn't that mean there are hundreds of cases where this needs to be fixed to resolve this bug?". The answer obviously being yes. Incoming 100+ file PR to resolve this issue. I have no other ideas for how someone is supposed to resolve an issue in this scenario
GitHub is big software, but not that big. Huge monorepos and big big diffs grind GitHub to a pulp.
Its these professional PM's that have done nothing else other than project mangement or PMP that don't have an understanding of the long term dev. cost of features that cause these systemic issues.
VDOM is also a good idea that simplifies the mental model tremendously. Of course these days we can do better than a VDOM. Svelte in fact doesn't use a VDOM. You can say that VDOM is a terrible idea in comparison with Svelte, but that's just anachronistic.
That’s one of my favorites. The exact bug they described during React launch presentation, that React was supposed to help fix with the unidirectional dataflow. You know the one where unread message badges were showing up inconsistently in the UI in different places. They never managed to fix that bug in the 10 years since React was announced and I eventually left Facebook for good.
should they be locking safari to the OS, definitely not. but users can just go download another browser if they are actually concerned.
I'd put it on the end user for not updating software on 15 y/o hardware and still expecting the outside world to interact cleanly.
It's really hard to fight the trend especially in larger orgs.
I'm still a big believer in "separation of powers" a la Scrum.
There should be a "Product Owner" that can be anyone really, and in the other side there is a self-managed development team that doesn't include this participant. This gives the team leverage to do things their way and act as a real engineering team.
The reason scrum was killed is because of PMs trying to get themselves into those teams and hijacking the process. Developers hated "PM-based scrum", which is not really scrum at all.
Which, unfortunately, cannot be measured :( so no KPIs. Darn!
Its all fun and games until you cut quality over and over so much your customers just leave. Ask Chrysler or GE. I mean they must have saved, what, billions across decades? And for free!
Well... um... not free actually, because those companies have been run into the ground, dragged through hell, revived, and then damned again.
A computer will be able to tell that the 497th has a misspelled `CusomerEmail` or that change 829 is a regexp failure that trimmed the boolean "CustomerEmailAddressed" to "CustomerEmailed" with 100% reliability; humans, not so much.
That's probably true.
> 15 y/0
It's a matter of expectations, many laptops that old still work decently enough with a refreshed battery. Funnily enough win10 was released 15 ago, and one can still get support for it for at least another 3 years until 2028, even on the customer license.
> Writing on the internet can be a two-way thing, a learning experience guided by iteration and feedback. I’ve learned some bad habits from Hacker News. I added Caveats sections to articles to make sure that nobody would take my points too broadly. I edited away asides and comments that were fun but would make articles less focused. I came to expect pedantic, judgmental feedback on everything I wrote, regardless of what it was.
https://macwright.com/2022/09/15/hacker-news
Which is true. Pedantism is the lowest form of pseudo-intelligence.
IMO "Knowing enough to do damage" is the worst possible situation.
A regular user who's a domain expert is 100x a better PO.
This is such a tired trope.
Sure 1000+ changes kills the soul, we're not good at that, but sometimes there's just no other decent choice.
Of course some languages... PHP... aren't so lucky. $customer->cusomerEmail? Good luck dealing with that critical in production, fuckheads!
Or that you had to avoid Ctrl+F "CustomerEmail" and see whether you had 1000 matches that matches the number of changed files or only 999 due to some typo.
Or using the web interface to filter by file type to batch your reviews.
Or...
Just that in none of those cases there is anything close to our memory/attention capacity.
Now you CAN so it so that is not the case, but tbh i have never seen that in the wild -
The reality is both can be slow, it depends on your data access patterns, network usage, and architecture.
But the other reality is that SPAs and REST APIs just usually have less optimal network usage and much worse data access patterns than traditional DB connected SSR monoliths. Same goes for micro service.
Like, you could design a highly scalable and optimal SPA. Who's doing it? Almost nobody.
No, instead they're making basically one endpoint per DB table, recreating SQL queries in client side memory, duplicating complex business logic on the front and back end, and sending 50 requests to load an dashboard.
I work in a large C++ codebase and a rename like that will actually just crash my vscode instance straight-up.
(There are good automated tools that make it straightforward to script up a repository-wide mutation like this however. But they still generate PRs that require human review; in the case of the one I used, it'd break the PR up into tranches of 50-ish files per tranche and then hunt down individuals with authority to review the root directory of the tranche and assign it to them. Quite useful!)
Edit: here's a good investigation on a real-enough app https://www.developerway.com/posts/tailwind-vs-linaria-perfo...
Even other frameworks like Vue.js, Solid or Svelte don't really suffer from it as much. It simply happens a couple order of magnitudes more often in React than any other framework.
Tailwind is probably one of the best considering you can use Vite to literally strip out all unused css easily.
And I think tailwind v4 does this automatically