Re: Cost analysis: (was: Getting to Consensus: CONTINUATION-related issues)

On 19 July 2014 04:24, Martin Thomson <> wrote:

> Did I miss anything else pertinent to the discussion?

Yes - the option c - namely the "drop continuations and fragment headers
proposal" in the wiki.

It avoids the role back issues of a).  Can work with either a compressed
header limit, an uncompressed header limit or no header limit. supports
streaming headers & unlimited headers.  It can be interleaved if we wish
and while that may slightly increase the DoS surface, it is at a big gain
in protection from HOL blocking and multiplexing quality of service, plus
it does not require interleaving support - just allows it.

I really think it was the best proposal that consider all the 'can't live
with' points raised on this subject.  But it really did not get a fair
consideration (not blaming the chair for that - I saw the same lack of
enthusiasm as he did in the WG).

It is a fine mess that we find ourselves in!


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

Received on Saturday, 19 July 2014 00:31:44 UTC