Hacker News Reader

Comment by TrackerFF | Hacker News Reader

Dunno if it's just me, but that website is more sluggish than flightradar24, which has hundreds of moving data-points on its map. What gives

View on HN

Or why it's attempting to make a bunch of post requests to example.com...

Seems like this isn't ready for primetime.

Almost unusably slow in firefox on a decent macbook pro, it looks nice though.

Probably has something to do with the fact that every movement is pushed to the browser history.

Try pushing back button after you have moved the map a little.

I doubt it has anything to do with that at all, that's not a performance intensive operation at all. Rendering a map, however, can be very performance intensive when done the wrong way, as I suspect it is being here.

I meant “something to do” as in it’s symptomatic of some repeated operation - not the direct cause.

Since each pushState also creates a object reference that cannot be garbage collected as normal (every pushState includes a object to be passed to a popstate event) I don't think it's too wild to suspect slowdowns to be caused by them using pushState instead of replaceState.

It could of course be thousands of other things too (and in combination), but it's not a bad place to start.

It's slow for me, on a 32 core Threadripper with a Titan RTX ;)

I usually don't notice slow websites, but it's noticeably blurry as I pan it around because it can't update at the 165Hz refresh rate of my monitor. Reminds me of gaming in 60Hz again.

I looked into it a little more. It switches into night mode when it gets dark out (or they pushed an update), and the performance is much better. My subjective impression was that it was usually getting a decent framerate when panning, but with some stuttering. I took a profile of it in Chrome, and that's what's going on. A lot of the time, it will run at 120-130Hz, but drop down to 50Hz for no real reason. (They sleep a bit between frames, they aren't just maxing out the CPU. Probably tested on a 60Hz monitor, and they don't play enough games to be able to pick out the drops to 50Hz. And I guess they didn't use the profiler, or look at the fps profile that's provided.)

They are doing some weird things. The Javascript console is littered with CORS warnings -- for making a cross-origin request to example.com every time a map tile needs to be loaded. That is probably an accident. They also seem to make HTTP requests for the map tiles whenever they enter the screen; they appears to be no caching whatsoever. (Maybe that's the right approach. The tiles do end up being served from the browser's cache. The network inspector claims they take about 1ms to load, which is probably slower than a lookup in a data structure in the Javascript program... but I didn't test it and I don't really trust that a "1ms" in there means "we timed it and it was exactly 1ms". It probably means "0 < time < 1", which could be anything.) Finally, they are serving this all out of what appears to be Azure blob storage; it's HTTP/1.1 so I guess we are opening a new TCP connection each time a tile needs to be loaded.

My TL;DR is I don't really do frontend stuff or performance stuff, but the site seemed especially slow. I bet they could clean it up in a few hours, so probably will.

Literally caused music playback to stutter on my iPhone. On desktop it was okay. Defiantly needs optimization for phones.

You must have an absolute beast of a desktop :)

I have a reasonably recent mid-range desktop and zooming in and out is impressively slow in Safari. It's even slower in Chrome.

It barely renders on my core i7 MacBook Pro. 32gb ram. Radeon card permanently enabled. Safari browser.

Yep, slow on a brand new macbook pro. Kinda unacceptable.

Reply on news.ycombinator.com