Comment by bambax
9 months ago
On recent hardware, how much "performance" do we really need? Wouldn't almost any compliant browser be basically good enough?
9 months ago
On recent hardware, how much "performance" do we really need? Wouldn't almost any compliant browser be basically good enough?
There's several decades-old sayings to the effect of what Intel giveth, Microsoft taketh away, or similar observations about the software side of computing spending all the hardware improvements and more.
To this general principle you can add browsers and websites; what the browser giveth, the websites taketh away. You may think browsers are slow... they really aren't! There's a staggering, even arguably insane, amount of optimization in there. But then we write websites that are barely adequate, and load them up with ad scripts that aren't even barely adequate, and blame the browsers for being slow.
Write yourself an old-school 1998-style static website without a big pile of fancy features, give yourself solid .css and .js caching and use it judiciously, and the browsers can blast content to the screen blazingly fast, for all the work it is doing.
If you even could feed a 2024 web site to a 1998 browser, you'd probably be able to eat a meal while it was trying to render facebook.
> You may think browsers are slow
I don't. I use uBlock Origin which blocks "ad scripts" and the like. My everyday machine is an old PC (older than 10 years) still on Win7, and everything is running just fine.
I also use a top of the line, recent PC on Ubuntu, mostly for development. Websites there feel instantaneous. I sometimes wonder what a subpar browser would feel like on that machine.
Maybe I should just try to run Ladybird on this to see how it goes.
So many apps have low-hanging fruit performance issues that don’t get addressed because they are judged to perform adequately in practice. Addressing them takes developer time, and not all developers have the skill set to do so (especially in a methodical way).
But, what if we had an AI agent dedicated to improving performance? It doesn’t need to be capable of solving every problem, but it could address the low-hanging fruit problems which aren’t hard to solve but nobody has time to look at.
As an embedded developer it always makes me sad to see physicists and engineers pushing the limits of physics to make faster hardware, just for devs to squander that power with lazy programming.
No, that's actually the point of faster chips: To make software development less challenging and cheaper.
Exactly. What people also miss that the complexity grew considerably because of the need to cover many more "edge" cases. 30 years ago, you could assume a rough display size (fixed layouts) and DPI (no scaling needed), assume ASCII / ISO-8859-1, assume that the user is able-bodied and doesn't need accessibility features, target just DOS etc.
There is also a lot of accidental complexity which you might be able to get rid of only by BC breaks, unfortunately.
1 reply →
No, it really isn't. The point of faster chips is for the user to be able to do more things and do them faster (without having to wait for the computer). This may mean more complex software that uses the added capabilities to do actually useful stuff that would not have been feasible before. It does not mean meaninglessly squandering performance on layers of abstractions because that takes slightly less developer time.
The point of faster chips is to make computer programs faster. We've increased performance a thousandfold and yet programs almost always take more than a frame (16ms) to update. Once everything is rendering at absolute 60FPS (and actually, 144FPS would be nice), and the hardware is reasonably cheap (Raspberry Pi cost perhaps), then the point is to make software development cheaper.
One would think so, but some browsers do not handle well repaints or do it prematurely. I've been testing a fediverse platform against a plethora of browsers, and I'm always surprised at the differences. It's not terrible, but some do take their time.
LadyBird author posted a couple of days ago a demo of twitter and he himself admitted that it's painfully slow.
To be fair, Twitter is painfully slow on other browsers as well due to only fetching and rendering the content after the page and javascript have loaded.
That sounds like a feature, not a bug.
As someone developing web games, my answer is no.
Web games are squarely outside "what do we really need" territory.
That's not up to you to decide.