Re: [quicwg/base-drafts] HTTP/3 references QUIC Stream IDs directly, allowing illegal references (#3273)

[Kazuho Oku <[e-mail]>]

While I agree that such a design is possible, I am in favor of using stream IDs as they are, because that is what we have done in HTTP/2.

To be precise, HTTP/2 has stream 0 (which is used for transmitting control messages), odd-numbered streams (for carrying requests), even-numbered streams (for carrying pushed requests).

To paraphrase, we have had these ID holes in HTTP/2, and we haven't had issues (I assume). We've already built our infrastructure around this design, for example, we simply log the stream ID and hope that that would be sufficient to identify the request, regardless of it being one initiated by the client or one being pushed by the server.

Introducing the concept of "request ID" would be a departure from this design in HTTP/2. As there would be a different number space for pushed requests, we'd be required to change the logging structure to convey a tuple of request-type and request-ID. While that is not impossible, I am not sure if that change is worth the cost.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/3273#issuecomment-557949331