RE: Proposal for various Infosetisms

Martin Gudgin writes:

>> I don't expect people to sign the envelope. 
>> I think people will sign individual headers, 
>> groups thereof or the body.

As an 80/20 cut, I agree completely, but why preclude it.  One could take 
the view that the simplest case is:  what I put in, I get out.  Now, if an 
intermediary really needs to put a header in, that's a change to the 
message and a signature should break.  If, however, we just want to make 
sure the message came out as sent (which doesn't preclude that headers 
showed up along the way as long as they were then stripped), then I think 
it's a really simple model to say that the infoset shows up intact, or at 
least that we don't specifically license unnecessary changes.

Conversely, if we go with your proposal, why would it not be more 
appropriate to make it symmetric and to say (roughly):  "A soapenv:Header 
with no children is semantically equivalent to providing no soapenv:Header 
element at all;  either representation may be used when sending a message 
with no header entries, and intermediaries are free to insert or remove 
such empty header element information items when relaying a message."

Why: Presumably your change is to allow an intermediary to build up some 
abstract list of headers, and to just use a canonical form when sending. I 
can think of at least as many cases in which the intermediary would want 
to unconditionally send the (possibly empty) element as to remove it. 

So, I think I'd either say "preserve it" (first choice, but not a strong 
preference) or "feel free to change it either way."  Remove only doesn't 
make as much sense to me.  Thanks.

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

Received on Tuesday, 1 October 2002 10:53:57 UTC