Re: Large Frame Proposal

On 8 July 2014 10:58, Mark Nottingham <> wrote:

> That's at most an editorial issue, and probably an invalid one; no where
> does the spec say you can ignore CONTINUATION, or that it's optional.

That is the issue!

Implementations are required to implement a modal feature that is only used
by 0.001% of traffic, with a switch between modes at an indeterminate size
and many implementations will never ever ever see any traffic that uses
this mode.

This is exactly like header field continuations in http/1 - a required
seldom implemented (correctly) feature used by infinitesimal users.

At the very least we should acknowledge that this is a problem, even if in
the end we accept it so we can handle big headers.


Greg Wilkins <> HTTP, SPDY, Websocket server and client that scales  advice and support for jetty and cometd.

Received on Tuesday, 8 July 2014 03:20:31 UTC