Re: Header Size? Was: Our Schedule

On 1 June 2014 02:21, Roberto Peon <grmocg@gmail.com> wrote:
>
>
> Poul's proposal to put routing info at the front pre-supposes that all
> proxies use the same information for routing.This is often not the case.
>

It is true that some proxies will want to look at arbitrary request meta
data - including cookies- to make their routing decisions.    Such proxies
are committed to full decoding.

However, there is also a large class of proxy that will want to look at
only the basic information.

So perhaps it is time that we restricted the transport meta data to only
carry transport meta data, and to move the extra meta-data to a secondary
header set?

Ie if we extended Poul's proposal so that all headers defined by the http2
specification for the use of transport, and only those headers, are in an
initial header set. A subsequent header set can then contain all of the
additional headers that an application/client/server may add.

Proxies that want to route only on the core http defined fields, can then
just decode the first header set.  Proxies that wish to route based on
additional headers (including cookies), would have to decode both sets.

I've always found it bizarre that HTTP design often results in allowing
applications to set headers like Transfer-Encoding!


This has been discussed before, I'll point out.
>

My limited experience of the IETF process indicates that WG such as this
often do have to re discuss design decisions made long ago.  This process
tends to iterate until either the spec and associated document because
clear enough to explain the main design decisions,  OR more likely that the
re-discussions appear to be mostly provoked from the same set of chronic
malcontents who wont accept they have not made their case.

For my own part, I have found the recent discussions very worthwhile and I
very much appreciate those who have taken the time to (re)explain why many
of the decisions have been made.   While I am critical of parts of the
spec, I am impressed at the depth of the reasoned explanations for almost
all of them.   So thanks for staying engaged and spending the time to
explain.

cheers



-- 
Greg Wilkins <gregw@intalio.com>
http://eclipse.org/jetty HTTP, SPDY, Websocket server and client that scales
http://www.webtide.com  advice and support for jetty and cometd.