Re: [quicwg/base-drafts] Semantics of SETTINGS_MAX_HEADER_LIST_SIZE (#2516)

[Kazuho Oku <[e-mail]>]

Yes. The problem here is the difficulty of the sender of the headers block complying to the limit advertised by the peer. Therefore, "SHOULD enforce" does not work as a fix.

@ianswett 
> In the case of proxies, it seems like the min of the received setting and local limit could be used? But I guess the proxy mah not know all backends it may contact when it advertises a limit. I still think something is better than nothing, but I'm not sure how to fully solve the proxy case.

FWIW, there are at least two issues regarding proxies.
* Downstream (e.g., origin server in case of a HTTP request) having smaller limits. In such case, the proxy will read the request then send a LIMIT_EXCEEDED error without forwarding the request.
* A proxy that streams the headers block. Such a proxy cannot tell the uncompressed size of the headers block until it receives the last byte of the HEADERS frame. But by the time that is observed, it would already have started sending the headers to the downstream.

-- 
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/2516#issuecomment-472617433