Re: i93: Repeating Single-value headers

fre 2007-12-28 klockan 18:36 +0100 skrev Julian Reschke:
> Henrik Nordström wrote:
> > fre 2007-12-28 klockan 17:19 +0100 skrev Julian Reschke:
> > 
> >> Hm. Do you really think it's SHOULD NOT, not MUST NOT?
> > 
> > My starting point is SHOULD NOT for anything which does not cause
> > immediate interoperability breakage.
> > 
> > If diskussions find that violating this is really harmful, not only
> > slightly confusing to the unsuspecting then upgrade the criteria to MUST
> > NOT.
> > 
> > It depends a bit on what the wording regarding error handling will be.
> 
> Well, if conflicting information in Content-Length headers doesn't cause 
> interop problems, what will?

Was thinking more on a general rule covering single-valued headers in
general, just to emphasis what the BNF says.


For Content-Length it should be a MUST NOT, but the more interesting
topic to discuss there is how recipients should react when receiving a
message with multiple content-length values.

Regards
Henrik