RE: NEW ISSUE: isolate TCP-specific aspects of HTTP

 

> -----Original Message-----
> From: ietf-http-wg-request@w3.org 
> [mailto:ietf-http-wg-request@w3.org] On Behalf Of Barry Leiba
> Sent: Monday, March 30, 2009 8:19 PM
> To: HTTP Working Group
> Subject: Re: NEW ISSUE: isolate TCP-specific aspects of HTTP
> 
> > To make the specification of such mechanisms easier and 
> less prone to error,
> > p1 should be arranged so that TCP-specific connection 
> details be isolated
> > from the rest of the specification. An alternate approach 
> would be to factor
> > such aspects into a separate draft, but this is probably 
> not necessary
> > (i.e., as long as it's easy to identify which parts of p1 
> are overridden,
> > that should be enough).
> 
> I strongly agree with doing this, and I agree that keeping it in the
> base doc is fine.  Something like this would work:
> 
> == Section [n] - TCP Considerations ==
> At the time of this writing, HTTP is used exclusively over TCP [ref].
> That need not be the case in future, and implementations SHOULD be
> adaptable to other transport-layer protocols underneath HTTP.
> 
> When TCP is used, the following apply:
> ...etc...
> 
> == end of Section [n] ==
> 
> Yes/no?

Can we say anything more than "SHOULD be adaptable"?  I don't know it might
mean.

Let's say, for example, my implementation supports running HTTP over SCTP --
what should my implementation do in order to meet that SHOULD?

-d

Received on Tuesday, 31 March 2009 16:44:40 UTC