RE: treatment of ns prefixes by intermediaries

> I assume that 'preserve' and 'reuse' are distinct,
> i.e., that an intermediary is not required to 'reuse'
> prefixes for inserted headers.

I think Matt raises an interesting question.  We have the model (which 
I've always thought was a bit too tricky, BTW), that processed headers are 
removed, but that possibly identical ones can be reinserted if the feature 
so directs.  Fine, but this means that each feature should specify whether 
the reinserted headers must preserve prefices. 

I think interop would be improved if we could basically say. 
"Specifications for such features SHOULD, when practical, ensure that the 
infoset of the reinserted header conforms to the rules that would have 
applied if the header had not been processed."  I'm sure there's a less 
lumpy way to say it, but I'm trying to give transparency to the next 
downstream node.  If A sends to B sends to C, then C gets the same header 
regardless of whether B ignored or processed and reinserted.  Possible 
spec clarification?

------------------------------------------------------------------
Noah Mendelsohn                              Voice: 1-617-693-4036
IBM Corporation                                Fax: 1-617-693-8676
One Rogers Street
Cambridge, MA 02142
------------------------------------------------------------------

Received on Tuesday, 11 February 2003 19:57:53 UTC