Re: HTTP/2 vs 1.1 semantics: intermediate codes

On Thu, Jun 12, 2014 at 10:48:10AM +0200, Julian Reschke wrote:
> On 2014-06-12 10:41, Willy Tarreau wrote:
> >On Thu, Jun 12, 2014 at 09:53:57AM +0200, Julian Reschke wrote:
> >>>Do you have a use case in mind where it could really make a difference
> >>>to have them ?
> >>
> >>I kind of liked the ideas behind 102 (which used the message for
> >>progress reporting).
> >
> >I don't remember it to be honnest, I'll have to take a look.
> >
> >>My preference would be not to break 1.1 features that aren't broken
> >>unless they clearly make HTTP/2 more complex. Is this the case here? At
> >>least we shouldn't claim we have a better replacement if we don't.
> >
> >I agree with that principle. At the same time I think that if we can do
> >without it's still better just to avoid carrying some of the 
> >interoperability
> >issues we had (eg: clients don't wait too long for 100-continue, 
> >intermediaries
> >have to consume all of them even if multiple responses are sent, etc).
> 
> But that's a problem specific to 100, right?

Yes clearly.

> I'm totally OK with that one being killed.

Then maybe you're right and we should allow 1xx to pass and redefine
the semantics there when codes are unknown from the client (ie:
currently 1xx cannot fall back to the same behaviour as 100 contrary
to other codes).

Willy

Received on Thursday, 12 June 2014 08:52:38 UTC