RE: Proposed Clarification for Issues 4 and 23

Andrew Layman writes:

> I'm not sure that there would be much complexity added by
> allowing processing instructions.  Parsing-wise, they are on the
> same level of ease as comments, which SOAP allows.[1]
Agreed.

> Semantically, my understanding is that processing instructions in
> XML generally are always safe to ignore except by applications
> that are coded specifically to recognize specific processing
> instructions.  So, in general, any application that is not coded
> to process a specific PI may ignore it.  Unless my understanding
> of PIs is mistaken, that should make for a fairly short
> discussion, much as Henrik proposed.
> 
I guess my concern is that there might be consequences for the stack as a 
whole.  Should WSDL be enhanced to indicate which PI's are legal and for 
what purpose?  If I do a processor and API that's designed to deal with 
SOAP in particular, I can today do things that are lots simpler than for 
general XML (e.g. no internal subset.)  Do I need to find a way to expose 
the PI to applications?  After all, someone might hand me that enhanced 
WSDL and tell me that their protocol depends on my giving them access to 
the PI.   If so, is it most likely to be associated with the SOAP header 
entry in which it appears?  If so, I give it to the associated software 
most likely, I think.  What it it's elsewhere in the envelope?  I 
understand that I can just say:  you can put it in, I won't say what it's 
associated with.  I'm not sure that, in a SOAP specification, that's 
desireable.

So, large or small, that's the sort of complexity I had in mind.  I think 
it needs to be justified vs. the benefits.

> PIs have some utility.  For example, one might put a stylesheet
> PI into a SOAP message so that it would look nicely rendered in a
> browser or email, and not just look like raw XML.  This could be
> helpful in getting more SOAP/XMLP adoption and also helpful in
> getting email messages to be transmitted as SOAP/XMLP
> messages. Just an idea.
I guess I still feel:  keep it simple unless the use case is compelling. 
If we do it, don't kludge it.  State clearly the range of uses that are to 
be supported, any intended associations (or lack thereof) with header 
entries, etc.  I'm not quite convinced that just "allowing" PI is saying 
enough to promote interop.

------------------------------------------------------------------------
Noah Mendelsohn                                    Voice: 1-617-693-4036
Lotus Development Corp.                            Fax: 1-617-693-8676
One Rogers Street
Cambridge, MA 02142
------------------------------------------------------------------------

Received on Monday, 25 June 2001 13:53:11 UTC