reverseproxy: ignore RFC 1521 parameters in mime type when checking the Content-Type header

[cmfcmf]

Without this change, a Content-Type header like "text/event-stream;charset=utf-8"
would not trigger the immediate flushing.

Fixes #3765

[CLAassistant]

)
All committers have signed the CLA.

[mholt]

Thanks for the PR! I will look at this soon.

... and I'm kinda beginning to wonder if -1 should be the default for everything at this point. Thoughts, @francislavoie / @mohammed90 ?

[francislavoie]

Idk, that sounds like too risky a change to me 🤔

[mholt]

Why?

[francislavoie]

Isn't there risk that we break behaviour for some clients? There must've been a good reason for it not to be set to -1 by default in stdlib, no?

The code comes from here https://golang.org/src/net/http/httputil/reverseproxy.go?h=FlushInterval#L410

Should we bubble this change upstream to see what they have to say? I know our proxy implementation is not dependent on it but it feels like we should let them know of fixes like this?

[mholt]

Isn't there risk that we break behaviour for some clients? There must've been a good reason for it not to be set to -1 by default in stdlib, no?

We don't know if it is likely to break anyone; the worst I imagine could happen would be a performance degradation. (I mean, I'm sure it could break someone but I doubt it'd be the majority case.)

Anyway, that's tangential, just thought I'd bring it up...

Should we bubble this change upstream to see what they have to say? I know our proxy implementation is not dependent on it but it feels like we should let them know of fixes like this?

Sure. Sounds like a good idea.

[mholt]

Thanks for the fix @cmfcmf !

[cmfcmf]

Should we bubble this change upstream to see what they have to say? I know our proxy implementation is not dependent on it but it feels like we should let them know of fixes like this?

Sure. Sounds like a good idea.

I have now opened an upstream issue about this fix: golang/go#47359