Re: Finding consensus on alt-svc, was: http://tools.ietf.org/html/draft-nottingham-httpbis-alt-svc as a normative reference in http/2

On Thu, 20 Mar 2014, Julian Reschke wrote:

>> It is correct that we are interested in using Upgrade and not Alt-Svc for 
>> upgrading HTTP URIs from HTTP 1.1 to HTTP/2.  I am not aware of
>
> My understanding is that Microsoft thinks that Upgrade on port 80 is going 
> to work well enough in practice that no other mechanism is needed. 
> Apparently, the other implementers do not believe this, or are opposed to 
> this for other reasons. It would be awesome if there were actual numbers on 
> success rates to inform the rest of us.

We support http1+Upgrade: on port 80 for plain-text HTTP2 in curl (powered by 
nghttp2) as well as the other mechanisms and I intend to offer (at least 
partial*) Alt-Svc support as well. It is just not there yet.

Primarily of course because we make a tool widely used for exploring, testing 
and debugging these internet protocols so supporting "everything" is what we 
do, but also because I'm a proponent of being able to switch HTTP:// traffic 
to use TLS in a controlled manner. (I'm avoiding the O* acronyms!)

There aren't that many http2-draft9/10 servers alive out there yet that 
support upgrade (hint hint) so I have no numbers or even a hunch of how good 
it actually works in real-life.

[*] = because the setting up of a secondary route to the same content is 
intended to be somewhat of an asynchronous operation it isn't that easily 
transferable to curl paradogms. It'll have to work on a more direct and 
explicit way for us.

-- 

  / daniel.haxx.se

Received on Thursday, 20 March 2014 07:39:17 UTC