Re: Getting to Consensus: CONTINUATION-related issues

In message <F38E04EB-CF59-451B-B080-7ADE7424F258@mnot.net>, Mark Nottingham wri
tes:

1)  Prefer a)

2)  No to b).

The lack of upfront indication of compressed header size puts the
DoS and memory management trouble where we least can afford it from
a performance point of view.

The only quasi-sensible argument for fragmented header transmission I
have heard pertains to server -> client communication at one particular
very well-funded website.

I have not seen a shred of evidence that "header-streaming" is
relevant to any request in client->server direction where our main
DoS/performance problems lie.

Imposing performance and DoS resistance costs on everybody else for
the sake of this one website makes no sense.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.

Received on Friday, 18 July 2014 05:59:25 UTC