CommonClient: handle Firefox removing the password part of userinfo

[Berserker66]

What is this fixing or adding?

Firefox's mistake
More details in https://discord.com/channels/731205301247803413/1201031114400210984

How was this tested?

by copying a link Firefox imagined into existence.

If this makes graphical changes, please attach screenshots.

[ThePhar]

Opened an appropriate bug report for Firefox: https://bugzilla.mozilla.org/show_bug.cgi?id=1876952 as well.

Believe we are not following spec based on report from Firefox. Withdrawing approval to review other solutions.

[black-sliver]

my suggestion would be to do

    if server_url.username:
        ctx.username = server_url.username
        ctx.password = server_url.password or ""

instead to meet the spec used by firefox/browsers.

[Berserker66]

my suggestion would be to do

    if server_url.username:
        ctx.username = server_url.username
        ctx.password = server_url.password or ""

instead to meet the spec used by firefox/browsers.

this will still crash when passed into the websockets library, which specifically checks for this and raises an exception. In the discord thread I suggested raising an issue with said library as another way.

[black-sliver]

this will still crash when passed into the websockets library

Hm, I don't have a good setup to test this without first implementing archipelago:// on linux, so I'll let you and phar decide, I guess.

[ThePhar]

Is this something we still want to handle like this? Unfortunately seems browsers want to follow a different URI standard than Python does.

[Berserker66]

I'd still prefer to do things technically correct, but browsers overwriting html with their own imagination is not something we can do much against.

[black-sliver]

It looks like Firefox undid their change. I have not checked other browsers. There seem to be a bunch of regressions with other URIs. https://bugzilla.mozilla.org/show_bug.cgi?id=1603699 seems to be the upstream "request" to break it (again). The last 4 posts are kinda interesting. Should we just wait and see where it goes? Should we "fix" python websockets?

[Berserker66]

I guess we'll need to detect Firefox user agent in the long run and serve them their own html, I'm sure they'll love that. :P

[black-sliver]

I think WHATWG is not Firefox-exclusive, so others may follow if they haven't already.