Re: Getting to Consensus on 1xx Status Codes (#535)

On 18 July 2014 10:12, Martin Thomson <martin.thomson@gmail.com> wrote:

> On 17 July 2014 16:59, Greg Wilkins <gregw@intalio.com> wrote:
> > So at least it looks like all the voices here can live with 100
> semantics.
>
> That's taking liberties.  I think that Mark will make that sort of
> assessment.  And I note that the number of participants is pretty low.
>

by "here" I meant in this thread.   I was just trying to focus in on what I
saw as the remaining differences.


> > I prefer keeping RFC723x semantic exactly.
>
> So you've said.


Well no.   I've previously said I see use-case for the 100 semantic, but
not expressed a preference for a mechanism to transport it.      I'm
certainly not opposed to a new mechanism for the semantic, but I'm a bit
cautious about using the flow control window as the mechanism.


>  But you keep talking about mechanisms rather than
> semantics.  It seems to me like you want a near-perfect facsimile of
> the mechanism too.
>

No - I want the semantic.   But I see two proposals have been made for a
mechanism and I evaluated how they would affect my implementation.

As I already have code that implements the existing RFCs for 100 support in
h1, so it is simplest for me (and I expect a lot of others) to adapt that
to 100 handling for h2.  For jetty, most of the 100 handling is above the
framing layer and I have to explicitly disable it for non HTTP transports
such as SPDY, FCGI etc.

The jetty mechanism inspects request headers and sends a non final response
"frame" if the application signals it wants the body.    A h2 mechanism
similar to that, but that is incapable of carrying 1xx messages is fine by
me.  I just would like it at the same semantic level in the protocol.

A mechanism that uses framing flags and adjusting the flow control
algorithm to get similar but not exact semantics simply does not fit.   I
will end up implementing 2 similar but different 100 continues mechanism
depending on the transport layer - ugh!

regards
















-- 
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.