Hacker News Reader

The Thing is that most people here belittle Web Developers but they do not have a clue what we do. At all.

In your pretty short answer you’ve already caught mistake. Let‘s see:

> And if you want to stick the header to the top you can use `position: sticky` in CSS instead of including a huge Javascript file that can do all kinds of fancy stuff you don't need.

Your „simple trick“ to avoid „including a huge javascript file“ doesn‘t work in: Opera, Chrome for Android, IE and a couple more[1]. It does work in Safari, but not as you would expect. Also, position:sticky is still a Working Draft. It is suspect to be changed at all times. Want the same functionality across all Browsers? Better use Js.

Another thing is that most people here think „huge JS bundles“ are what make websites work. These are to be used for webapps though. If you want anything more than a static site, you‘ll need JavaScript. No way around it.

[1]: https://caniuse.com/?search=Sticky

> doesn‘t work in: Opera, Chrome for Android, IE and a couple more[1]

Are we looking at the same compatibility tables? Caniuse lists it as working in android chrome and opera with the exception of <thead> elements, which are not relevant here because OP's example was about <h1>, not <thead>. I'll give you IE, but that's already ignored by many.

> If you want anything more than a static site, you‘ll need JavaScript.

Forms are still a thing. Then there's hover animations, summary/details, video elements and a few other interactive things that don't need javascript.

There also is a middle ground between "static page" and "huge JS bundle". A little javascript usually isn't what makes web pages slow.

Come on, even the caniuse page you linked says `position:sticky` is supported by 94.33% of browser. For that ~5% left (very old browser versions or specific cases like Opera Mini, most likely even less real cases due to bot noise in statistics) there is a thing called "Graceful Degradation". If you do things right your page won't break at all if that CSS property is not supported. Your header will just be positioned fixed or even static.

So, really, no huge (or small) JavaScript file needed at all in this case. Like most other bloat in web dev, that's just laziness that accumulates.

94% is actually a really low number for support.

that means that 6 out of every 100 people can't use it.

Roughly 10 million people visit HN every month.

that means that about 600,000 people can't use it just on HN alone.

Well, statistics don't work in that linear way. For a start, that 94% is a global percentage that isn't evenly distributed worldwide, let alone on HN...

Anyway in my comment I was actually saying that if you do things right it will work even for that 5%, without any JS. Maybe it wasn't clear enough?

Flip it around: You're fucking up the browsing experience of 9.4 million people for a minor graphical effect that totally destroys scroll behavior and CPU.

Just use CSS. No one will miss the sticky bar if it doesn't show up on their browser.

> If you want anything more than a static site, you‘ll need JavaScript.

Who doesn't want static sites? I certainly do. They're faster and easier to use. I've never heard a user say "I wish my back button didn't work and the page rendered wonky so that buttons move as I try to click them".

It's web developers that push to turn simple HTML documents into "applications".

Every now and then I wish browser writers would compare timestamps between the link click and the last relayout, and throw away stale clicks instead of navigating to who knows where.

> If you want anything more than a static site

The problem, as always, is (poor) developers making poor decisions. Implementing static sites as a JS-required, JS-heavy pages is something that some poor development teams are doing.

Bungie, game developer, have been replatforming their site and they've turned what is mostly static content pages into purely client-side single page app which requires javascript to display anything https://www.bungie.net/7/en/Destiny/BeyondLight It's sad because there are extremely productive ways to build this out using their preferrred technoligies (react) but deliver the static content from the server in the initial page load AND continue doing SPA-style navigations if desired.

---

There are lots of webpages out there, that means there's lots of developers making them. Some developers will be good, some will be average, and some will be bad.

The 'problem' with websites is that it's a lot easier to see when they've been done poorly, compared to say native apps.

"Want the same functionality across all Browsers? Better use Js."

This is not the point. The point is some developers just include big Javascript files because they don't know what other methods and options there are.

I have seen developers include Javascript to 'fix' things across browsers but it made things worse because the `let` keyword was not supported in some browsers at that time.

So it is not about if you should use Javascript. It is about the fact a website or webapp can and should be fast. But when a developer does not know about the technical aspects we will end up with multi-megabyte pages that take 10 seconds to complete.

--

About the `position: sticky`: it is not smart to promise a client a pixel perfect experience across all browsers (unless they want or need to and will pay the extra price). In the case of a sticky header you can support all modern browsers with two lines of CSS (`position` and `top`). The other 6% won't have a sticky header but still a perfectly working website. So imho you did not caught a mistake but a real world example.

> If you want anything more than a static site, you‘ll need JavaScript. No way around it.

I’ve been noticing this kind of rebranding of what static means a lot. It seems to create lots of confusion (you can see it in this thread already).

Traditionally, “static” websites were websites without a server-side programming language backing them—sites whose content was unchanged by any users. Javascript, CSS, and HTML are the tools used to build static websites.

Today, some developers seem to use static to mean Javascriptless.

The term fits both situations pretty well, so I get why it happened. I don’t know if there’s a better less-ambiguous term used for either of these things now?! Non-database-backed?! Nonjavascript’d?

The term for JS enabled pages used to be DHTML back in the day. You'd be talking about a static-server DHTML site. Or something like that. https://en.m.wikipedia.org/wiki/Dynamic_HTML

Good point - Hacker News works with barely any javascript (and it's optional); hard to argue that a forum is "static"!

I'm no expert, but I hope someone at quora has some sort of light-bulb moment reading this....

Looked at the source for one answer because of their login wall and couldn't be bothered to re-create my account and to get to the actual answer is about 30!! nested divs and then each line of text (provided you've returned to the next line in the visual text editor to write your answer)is in its own <p> which is fine on its own, but each <p> is styled with a boatload of classes for no apparent reason....

I'm fully with you as far as loading times, but even a simple wordpress install these days add incredible useless stuff that makes it non trivial for the average blogger to simplify...

I think there are a couple of reasons to this:

1. Lack of expertise/interest, as you say. One root cause here is that there are many devs in web development who are self-taught, with no CS background.

2. Time pressure and contractual agreements: Things need to be finish ASAP to reduce cost, and implementation contracts usually state features and looks, but not performance or maintainability.

3. Many effects, such as sticky elements, couldn't be done in CSS just a few years ago, so devs got used to include libraries for these things. In fact, I think many modern CSS features got introduce because they were widely used, but with JS implementations.

> there are many devs in web development who are self-taught, with no CS background.

That's an excuse when you're just getting started, but after 3 or 4 years everyone has had time to figure these things out. There's papers, blog posts, youtube videos... even wikipedia can help in understanding performance on the web. If you've been doing web for 5 years and still don't know these things, it's not because you're self-taught; it's because you're lazy and/or being overworked with not enough room for learning.

> But I am not sure I can blame those front-end developers.

Also in a lot of cases you can build the fastest website with minimal markup and only the necessary CSS/JS libraries for the base functionality but if then for "business reasons" you need ad networks, google tag manager, analytics scripts, third party support popups there's only so much you can do as a front end developer to keep the site fast.

This all sounds true, but it breaks down about the fact that bare html and css are too low-level and quirk-rich for a high-level programming under reasonable budgets.

Also, not programmers are guilty of such bloat, but the tools that they use, html css included. There is no compiler that takes an idea / a representation of it and turns into highly compressed, DCE'd, -O3 LTO'ed bundle. There is some progress in this direction, but it is not that great and requires an effort to use.

because Bootstrap told them to do so, they nest as much DOM as possible

Bootstrap also does that for reasons. For reasons that are unclear to bootstrap itself, because these all lie in the inconsistencies, legacy-ness and incompatibilities of the allegedly most compatible platform ever. And when we try to discuss how to extend it (not even replace!), someone brings idealistic arguments about how it should stay as is, because otherwise (again allegedly) they couldn't print a webapp or screenread an html5 platformer game.

HTML needs an <include> tag to import HTML files (not full documents in general, just to reuse common pieces)

Isn't that roughly what server side includes are? https://en.wikipedia.org/wiki/Server_Side_Includes

To me not really, we do not use server side includes for JS or CSS, there are use cases where HTML would also profit from client-side imports

It sounds like what is needed is an `optimising compiler' for the entire front end that would spot patterns like those you mention and rewrite them into something better, and that would tree-shake out unused stuff.

Does anything like that exist?

That seems like it's going to work in very specific situations only and/or be extremely hard to implement.

I'm curious if Svelte is headed in this direction.

That strikes me as a heavy handed option given the prevalence of linters and performance tools which can already detect some of these issues.

I work for a company dominated by non CS . They brought in contractor since the contractor's CEO is a good talk. This contractor is full of full-stack JS devs. Now we have their ideas implemented on edge devices. Noice.

> ideas implemented on edge devices. Noice.

What does that mean?

Edge devices here means the most modern browsers?

And noice means unnecessary?

Edge devices are usually those operating on the edge of your network, topologically speaking (e.g. end-user devices which only have one connection to the network).

Noice is just an awkwardly phoneticized spelling of the word nice intended to convey a sarcastic tone.