The Web Services Architecture WG position on XML profiling/subset ting

 This email was originally sent to a limited distribution, and is being 
sent to the www-archive for public access.

Noah


                 To: "Champion, Mike" <Mike.Champion@SoftwareAG-USA.com>
                 From: "Noah Mendelsohn" <Noah_Mendelsohn@us.ibm.com>
                 cc: "''C. M. Sperberg-McQueen' '" <cmsmcq@acm.org>, 
"'David Orchard '" <dorchard@bea.com>, "'fallside@us.ibm.com '" 
<fallside@us.ibm.com>, "Champion, Mike" <Mike.Champion@SoftwareAG-USA.com>
                 Subject: RE: The Web Services Architecture WG position on 
XML profiling/su  bset ting


This note is a collective response to some of the replies I've received to 
the Tag note posted at [1].  I don't want to belabor this, but I've gotten 
some well-considered replies, all of which were either private to me or 
sent to this cc: list.  None were sent to Tag.  I've learned some things 
from these responses, and they've somewhat clarified my thinking.  So, at 
the risk of unnecessarily dragging this on, here's a clarification of my 
original concern, and a response to at least some of the specific points 
I've seen raised in feedback.  BTW:  I have no problem at all with this 
note or my others being forwarded to the Tag list, as long as enough is 
forwarded to give accurate context.

First of all, Mike hints that I may not have read all of his note 
carefully enough.  As he suggests, I don't think that's the whole problem 
or even most of it, but it may have contributed some.  So, apologies if 
that's happened. 

Clarification of my Original Concern
------------------------------------

I realized as a result of recent correspondence that I have some 
assumptions about the Tag and W3C process that others may not entirely 
share.  I view the Tag as essentially a senior executive body in W3C.  I 
see it's role as being to help the Workgroups succeed at what they do, and 
very occasionally to directly tackle certain issues with scope so broad 
and architectural that doing it "at the top" seems to make sense (what a 
URI references strikes me as one of those issues that it's reasonable for 
the Tag to discuss directly.)  I see it's role as being like the CTO or 
associated executive technical committee in a corporation.

I and others have benefitted from that fact that the Tag seems to welcome 
informal input and discussion from almost anyone in the community.  I 
think that's great.  On the other hand, I think one wants a much more 
formal relationship when a workgroup provides formal input to the Tag.   I 
think that in such cases the workgroup should have a responsibility to 
proceed cautiously, check pertinent facts, and within reason coordinate 
with other workgroups that may already have explored the pertinent issues. 
 Consider a technical team in a large company dealing with its senior 
management or CTO.  Sure, they may talk informally in the hall, but if the 
team approaches the CTO or his/her technical committee and says "we have 
formal input for you on an issue", then it's assumed that they've done all 
reasonable coordination and research on the topic before going "up that 
high".  The essence of my concern was a perception that in failing to do 
this, WSA actually misinformed the TAG about several points of fact (see 
below)
.   I thought it also implied to the TAG a consensus in the WSA/SOAP 
community that was in fact never explored outside of WSA.    Maybe some 
confusion comes because others don't share my view of how the Tag/WG 
relationship should work, or maybe it's because WSA though they were being 
sufficiently careful and I didn't see it.  Anyway, that was my concern.

I infer that WSA may think that the substantial membership overlap with 
XMLP and WSDL WGs mitigates the need for formal coordination.  I disagree. 
 The fact that there is some shared membership between the groups is a 
good thing...but it means nothing officially.   In W3C every member of a 
WG has equal standing and should have equal opportunity to engange in 
coordination activites with other groups.  I guess I have particularly 
high expectations for WSA, which is in some sense a mini-TAG for the WS 
area.  It whould be particular careful to work with and through the other 
groups in the WS area, I think.

Anyway, I think that's the big picture as I see it.  Rightly or wrongly, 
that's where my concerns came from.  I don't propose to pursue them 
further unless the Tag finds that helpful (if more discussion springs up 
on the Tag list...I may forward this note as a clarification of my intent. 
 Given that things are quiet, I'd prefer to let it go.  DaveO:  you 
certainly have my permission to forward this note to other Tag members if 
you feel that would be helpful.)

Responses to Particular Points
------------------------------

This is a long note already, but I thought it might be helpful to list the 
particular points I think I've seen in various responses, and to give a 
quick reaction.  I've done this without attribution:

>> You might not have read the original WSA note carefully.
As noted above, possibly true.  If so, I apologize.  I might change some 
emphasis in my note, but I think the overall concerns above remain for me.

>> Yeah! this needed saying, thank you!
You're welcome.

>> You're off base, and your tone was inappropriate.
I hope not, but if so I do apologize.  It was not my intention to be 
inflamatory.

>> Asking groups to check with each other before providing fomral input to 
TAG adds too much process overhead.
I respectfully disagree.  I'm not talking about an n-squared process here. 
 The WSA note specifically made proposals about SOAP and Schema.  In each 
case, WSA was commenting on areas that had been discussed in depth by the 
respective workgroups over a period of time.  I think that WSA would have 
done the Tag a much greater service by ping'ing the WGs first, and either 
arranging for joint approach to the Tag by the WGs together (we've gotten 
together, learned from each others' previous work, and at least can 
outline the alternatives) or failing that to at least say "we can tell you 
not only that there is division in our WSA group, but that the schema WG 
has found the question of schema subsets to be a subtle one involving 
difficult tradeoffs of simplicity vs. interop...we in WSA have not done a 
full analysis of those tradeoffs" (or whatevery you believe to be the 
correct summary.)

>> If WSA missed a technical point on SOAP, it must be subtle.  This is 
(maybe) proven by the fact the significant cross membership between the 
teams didn't serve to get it right.
You be the judge.  The SOAP CR [1] has in its table of contents:


            3.1.5.1 id Attribute Information Item
            3.1.5.2 ref Attribute Information Item
            3.1.5.3 Constraints on id and ref attribute information items

The first two of these merely state that the respective attributes are 
allowed.  The entire text of the third section is:

"The value of a ref attribute information item MUST also be the value of 
exactly one id attribute information item.  A ref attribute information 
item and an id attribute information item MUST NOT appear on the same 
element information item." 

The WSA note says that "One issue that SOAP 1.2 "punts" on (as do some 
other W3C specifications, including DOM Level 2 and XSLT) is the question 
of how an "id" attribute is defined in the absence of a DTD and without 
insisting on a schema processing step."

I don't think this is subtle.  I think the WSA group missed it, and if 
anything this demonstrates some of the risks in assuming that 
cross-membership is an appropriate foundation for cross group 
coordination.  It also suggests that members of the WSA did not look hard 
at the SOAP rec in making their proposal.  For an informal note among team 
members, I think this is fine.  It happens all the time.  As I said in my 
intro I think it's constructive to go for a higher standard when the Tag 
receives formal input from a workgroup.  My assumption is that the Tag is 
indeed too busy to read all the specs themselves.  I therefore fear that 
the TAG might take too seriously when a seemingly-authoritative source 
educates the Tag that "the XMLP group 'punted'".  This was one of the 
sources of the somewhat pointed tone of my note, and of the request for a 
formal "withdrawal."

>> The original WSA note and some later responses refer to a SOAP profile.
FWIW, I think the XMLP response (which admittedly I had a big hand in 
drafting) goes to some length to indicate that XMLP is not defining a 
profile of XML in SOAP.  I don't think I'm defining an XML profile when I 
invent an XML vocabulary for employee records, and I don't think we did in 
SOAP either.  Thus, even though WSA was going to some trouble to endorse 
the findings by XMLP, it then in a sense misquoted them.  That also upset 
me.  It might also lead the TAG to mistakenly believe that XMLP has 
consenus that it defined a profile.  I do agree that the constraints 
imposed by SOAP on its use of XML might be inspiration for a profile, were 
an XML profile to be a worthy goal.  Again, if WSA had first checked with 
the very group for which it helps set architecture, it would have 
uncovered some of these sensitivites...certainly from me, and perhaps from 
others.  I think the input to the Tag would have benefitted.

>> We need a profile because generic tools may not produce the subset of 
XML that SOAP uses. 
I am less convinced than others that SOAP's experiences are much of a 
justification for a subset profile, but that's a good discussion that 
should be aired.  It's surely not consensus from XMLP (since I am on XMLP 
and I would disagree at this point).  If it's consensus from WSA then it's 
fine for WSA to say so to TAG.  My concern is the one just above;  SOAP 
itself does not define a profile IMO.  One might or might not decide that 
it's inspiration for a profile. The WSA note implied that SOAP had already 
defined a profile that might be used by others.  The XMLP group never 
considered that question one way or the other, as I remember.

>> The Tag needs to do this work and WSA is just trying to help.
Actually, as I responded privately to Dave, I think the XML subset 
question belongs in the XML core workgroup.  I strongly disagree with Tag 
taking over design responsibilities from WGs that have been chartered in 
one or another area by the W3C membership.  I think it is appopriate for 
Tag to do technical fact finding, which in part is what they're doing.  I 
would prefer for the subset issue to relatively rapidly sent to core.  If 
core has doubts about working on this, then Tag should use whatever 
mechanisms are provided in the W3C process to ask the membership to 
clarify core's charter and responsibilities.  If feel the same way about 
xml:id, by the way.

>> You should have responded to WSA, not TAG.
As noted above, my concern was (1) that the TAG had been misinformed in 
several respects and (2) that this resulted from a breakdown in process 
that was in the end the Tag's responsibility (I.e. to decide what sort of 
work it wants its WGs to do before providing formal input.)  If the WSA 
note had not itself already gone to the TAG, I would certainly have 
preferred to come to consensus with WSA first.  Rightly or wrongly, I felt 
it as important to communicate my concerns to the Tag and tag list readers 
as to WSA.

I hope you'll forgive the long spiel above.  I really would rather not 
churn on this much more.  Indeed, I am essentially incommunicado Tues-Fri, 
which means I'll only be doing top-prioirty email until middle of next 
week.  So, I'd be glad to let this go.  You and the Tag have my input, and 
I hope it's helpful.  You also have my apologies for anything that is 
either in error or unduly inflamatory.  I would encourage the TAG and the 
WG chairs to think about the process issues.  Thank you. 

[1] http://lists.w3.org/Archives/Public/www-tag/2003Jan/0216.html
[2] http://www.w3.org/TR/2002/CR-soap12-part2-20021219/

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

Received on Monday, 20 January 2003 18:47:38 UTC