Re: Consequences of removing the reference set

On Fri, Jul 25, 2014 at 09:29:24AM +1000, Greg Wilkins wrote:
> On 25 July 2014 01:07, Martin Thomson <martin.thomson@gmail.com> wrote:
> 
> > * change 'header set' to 'header list'
> 
> 
> This is a good move.  Judging by the existence of duplicate content-type
> headers in the example data set then having an ordering to resolve is
> good.   Probably need to now say that if a header that cannot have multiple
> values is repeated then which value do we use, the first, the last,
> undefined or 400 ?

We had this issue already in H/1 and since H/2 must keep the same semantics,
I think we won't really address this case as it's implementation-dependant.
We processed the special case of content-length since it was a major issue
for the framing, but that's all.

BTW, good job Martin, the spec really looks great and simpler now.

Willy

Received on Friday, 25 July 2014 05:54:56 UTC