workbench.api.js takes very long time to load in 3.1.1 and not cached properly on FireFox

[marty1885]

Hi,

I've just updated code-server to 3.1.1. And I run my code-server on a remote server. I find the new code-sever taking fairly long (about 1.3 minute in my case). It seems to be caused by the long loading time of workbench.api.js. Furthermore on FireFox, this file although cached by FireFox; FireFox won't use the cache on the next load. Rendering opening code-server a waiting game.

Also, as a side note. My PC can connect to the server at ~30Mbits/s but workbench.api.js loads ~880Kbits/s. Which is quite slow.

(loading time of workbench.api.js on Chrome)

(FireFox loads the file every time)

Best

OS: Arch Linux AArch64
VSCode: 3.1.1
Installed via PKGBUILD

[code-asher]

Maybe we should gzip any file that's over a certain size or just explicitly gzip the workbench file.

You wouldn't happen to have "disable cache" checked in the Firefox dev tools, do you? (It's along the top near the Filter URLs text box.)

[marty1885]

No, I don't have cache disabled. (Otherwise I don't think it'd show up in FF's cached file list)

[code-asher]

I'm not sure; the checkbox might just prevent Firefox from using the cache rather than preventing it from storing into the cache. Could you verify whether the request has a Cache-Control header set? That's what we use to cache files so maybe it's not being sent somehow.

[marty1885]

My bad. FireFox sends Cache-Control: max-age=0 when reloading. Forcing to reload everything.. Is there a way around that? Maybe send 304 Content Not Changed?

[nhooyr]

That's interesting, why would Firefox do that?

[marty1885]

I believe it's the default behaviour from the older ages of the web when the only reason to reload a page is because something is broken so the cache can't be trusted. And this behaviour seems to be the compliant one with RFCs.

Reference:
https://stackoverflow.com/questions/60902008/why-does-firefox-ignore-cache-headers-and-revalidate-on-refresh

[code-asher]

Ahh, that's very interesting. I tested it out and going to the website again (typing it out, hitting a bookmark or a link, or hitting enter again in the URL bar) uses the cache but actually refreshing (using the button, F5, ctrl+r, etc) will not use the cache.

[code-asher]

Fixed by #2680.