Re: protocol negotiation

FWIW, the IETF Hybi work [1] on Websockets is proposing something very similar, 
but while it's HTTP-like it not exactly HTTP.  Sigh.

#g
--

[1] http://tools.ietf.org/html/draft-hixie-thewebsocketprotocol
     http://tools.ietf.org/wg/hybi/

Danny Ayers wrote:
> Henry, while I see nothing wrong with using a link, something like a 
> protocol change does feel like it should be a bit lower down the stack - 
> in fact just as Graham suggests (a header I can't remember seeing before 
> - really must read the manual sometimes).
> 
> On 28 February 2010 19:23, Graham Klyne <GK-lists@ninebynine.org 
> <mailto:GK-lists@ninebynine.org>> wrote:
> 
>     I may be getting this all wrong, but HTTP upgrade?
> 
>     http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.42
> 
> 
>  
> "Upgrade" sounds a bit strong, but that's exactly the kind of thing I 
> had in mind.
> 
> Just to clarify (my post was a bit late-night), I was imagining the 
> scenario where you have two agents/services wishing to talk with each 
> other, and http would be enough to do the identifiers and initiate 
> comms, but assuming other protocols were available. xmpp being a good 
> example, in the extreme case the agents/services might be running in the 
> same VM so direct method calls might even be in scope.
> 
> As an intermediate thing between such protocols, the recent work around 
> Activity Streams (http://activitystrea.ms/)  is quite interesting - big 
> crossover with RDF, the model is being reinvented mostly done using the 
> Atom format. I could imagine a bit of XSLT/XQuery in the pipeline were 
> it to connect with a triplestore.
> 
> Cheers,
> Danny.
> 
> 
> 
> 
> 
> 
> -- 
> http://danny.ayers.name
>

Received on Monday, 1 March 2010 19:05:44 UTC