RE: [w3c sml] [4775] Change "skip" to "lax" processing

All
Yet another attempt to instigate consensus.  This issue has been discussed threadbare in this thread ;-)

We have a proposal from Sandy and Kirk to change all occurrences of "skip" to "lax" in the SML IF schema  except the occurrence on DataType (which will continue to use "skip").

Please speak up now if you disagree with Sandy and Kirk's proposal - silence will be treated as consent :)

Thanks!
Pratul

From: Pratul Dublish
Sent: Thursday, September 06, 2007 8:27 AM
To: 'Wilson, Kirk D'; Sandy Gao; John Arwe
Cc: public-sml@w3.org
Subject: RE: [w3c sml] [4775] Change "skip" to "lax" processing

Kirk
Thanks for  clearly  articulating the two positions.  I have a few  comments and a question:


1.       "lax" becomes "strict" if the schema can be located

2.       SMLIF spec says nothing about the syntax/semantics of the extension points, and the extension points are irrelevant to the validity/invalidity of an SML IF document. Therefore, why impose the burden of attempting to locate the schema and validate them on all processors?

3.       A consumer who wants to validate the extension points can always do so even if the SML IF schema specifies "skip". E.g., this can be done by "extracting" the extension points from an SMLIF doc and validating them

4.       Question: Where is the apparent industry "best practice" re the use of "lax" documented?

Pratul

From: Wilson, Kirk D [mailto:Kirk.Wilson@ca.com]
Sent: Thursday, September 06, 2007 6:19 AM
To: Pratul Dublish; Sandy Gao; John Arwe
Cc: public-sml@w3.org
Subject: RE: [w3c sml] [4775] Change "skip" to "lax" processing

I support Sandy's fully articulated position, namely, that only DataType should have processContents="skip" for its content, and all other extension points (content and attributes) should have "lax" processing.

Pratul, I don't think the parallel between not worrying about the validity (or invalidity) of the documents and not worrying about the validity of the extension points is a good one.  Presumably extensions points are included for a reason and that reason probably has something to do with the intended processing of the SML-IF document.  Sandy's point that processors should enforce the SML IF structure does not imply what you seem to be saying, that they is all an SML IF processor should do-note Sandy's parenthetical comment.  IMO, those processors that are capable of understanding the extension points should be able to validate them.  Given the ubiquity of processContents="lax" in all other industry standards, I don't think it is much of a concern that all consumers that encounter extension points are forced to try to validate them.  (As I understand it, if the schema can't be located, then "lax" becomes in effect a "skip"-i.e., the extension point is not declared to be invalid.)

Just trying to keep the discussion going.  We seemed to have crystallized on two positions:

 1.  Sandy's position as articulated below (and in the first sentence above), which follows apparent industry "best practices", but with sensitivity to the special needs of SMIL-IF
 2.  The "pure performance option": maximize performance by skipping validation on every extension point.

Kirk Wilson, Ph.D.
Research Staff Member
CA Labs
603 823-7146

________________________________
From: public-sml-request@w3.org [mailto:public-sml-request@w3.org] On Behalf Of Pratul Dublish
Sent: Wednesday, September 05, 2007 5:21 PM
To: Sandy Gao; John Arwe
Cc: public-sml@w3.org
Subject: RE: [w3c sml] [4775] Change "skip" to "lax" processing

I agree with Sandy that SML IF schema should enforce the SML IF structure and not worry about the validity (or invalidity) of the documents contained in an SML IF document.  IMO, the same logic should be applied to the extension points since the extension points are provided for extensibility but are irrelevant to the structure of the SML IF document.  My understanding of processContents="lax" is that processors will attempt to find the schema and, if successful,  perform validation. Therefore, lax validation on extension points will require all consumers (or more precisely the XML Schema processor used by consumers) to attempt to locate the schema for extension points and validate them. In fact, a producer who uses extension points can force all consumers to validate them by including the extension point schemas in the IF document.  So, we should retain skip processing for the extension points in SML IF schema.

From: public-sml-request@w3.org [mailto:public-sml-request@w3.org] On Behalf Of Sandy Gao
Sent: Monday, August 27, 2007 6:47 AM
To: John Arwe
Cc: public-sml@w3.org
Subject: Re: [w3c sml] [4775] Change "skip" to "lax" processing


I agree with Kirk that this is (partially) a performance issue, because "lax" allows/requires (1.0/1.1) the processor to try to assess the entire subtree, whereas "skip" says "do nothing".

But there is a deeper issue here: what do we want the SML-IF schema to enforce? I think the answer should be to make sure the document satisfy the SML-IF *structure* (and any additional contracts/extensions between processors). That is, if a document being transmitted is invalid, it should *not* be a violation of the SML-IF schema. The IF is OK in this case. (Just like "The Moon is bigger then the Sun" is OK English-wise.)

What this means is that whether it should be lax or skip depends on what the wildcard is supposed to match:

- If it's for *extension" points (so that additional information can be attached to the SML-IF instance, to be interpreted by processors who understand it), then "lax" should be used, in case the processor has a schema that can provide components to validate the matching elements/attributes.

- If it's a place-holder for the document being transmitted, then "skip" should be used, so that we don't let validity of individual document to affect the overall IF validity.

Based on this, it seems that only "DataType" needs a "skip" wildcard for its content (not attribute), and all the others should be "lax".

BTW, why did "DataType" have, as its content:

      <xs:any namespace="##other" processContents="skip" minOccurs="0" maxOccurs="unbounded"/>

I would think it should be

      <xs:any processContents="skip"/>
      <xs:any namespace="##other" processContents="lax" minOccurs="0" maxOccurs="unbounded"/>

That is, we expect the first element to be the document being transmitted, which can have any namespace (including that for SML-IF). This one is "skip" because we don't care about its validity. This element must appear once and only once. Then there are any numbers of additional elements that can be used for extension purposes, hence "##other" and "lax".

Thanks,
Sandy Gao
XML Technologies, IBM Canada
Editor, W3C XML Schema WG<http://www.w3.org/XML/Schema/>
Member, W3C SML WG<http://www.w3.org/XML/SML/>
(1-905) 413-3255 T/L 969-3255
John Arwe <johnarwe@us.ibm.com>
Sent by: public-sml-request@w3.org

2007-08-24 12:01 PM

To

<public-sml@w3.org>

cc



Subject

Re: [w3c sml] [4775] Change "skip" to "lax" processing











yeah, what he said (+1 from me)

I never understood why we would prevent a validator from using schema components it could locate (skip), as long as they are not required (lax).

Best Regards, John

Street address: 2455 South Road, Poughkeepsie, NY USA 12601
Voice: 1+845-435-9470      Fax: 1+845-432-9787
"Wilson, Kirk D" <Kirk.Wilson@ca.com>
Sent by: public-sml-request@w3.org

08/24/2007 10:49 AM


To

<public-sml@w3.org>

cc



Subject

[w3c sml] [4775] Change "skip" to "lax" processing











All,

The email will serve to initiate the discussion of whether we should specify processControl="lax" rather than the current "skip" for wildcards in the SML XML Schemas.

There is only one occurrence of processControl="skip" in the SML specification: for the content of the smlerr:errorDataType.

The original issue was raised with respect SML-IF in which processControl="skip" is used for all extension elements (both xs:any and xs:anyAttribute) in the type definitions of this specification.

Since I wasn't involved in the original authorship of the spec, I'm not sure what the rationale was for the original use of "skip".  I assume it was for efficiency of the SML-IF consumer, the assumption being that the SML-IF would need to concern itself only with sml elements according to the semantics specified in SML-IF.  In my notes I have found the following definition of an SML-IF consumer: "processes SML-IF documents in whole or in part by the semantics of this specification" (emphasis added).  Since, by definition, extension elements lie beyond the semantics of the spec, there appears to be no reason for the processor's attempting to validate the extension elements.  But I would consider this a poor argument.

"Skip" seems too finalistic and may not meet the requirements of SML-IF consumer creators and SML-IF document authors who need to build in special information, eg., into the ModelType, and can provide the schema for validation (assessment).  I suspect that something like this rationale underlies what appears to be the industry "best practice" of using "lax" processing.   The cost of using lax processing is undoubtedly absolutely minimal.

I will recommend changing the spec to "lax", according to what is industry best practice.

Kirk Wilson, Ph.D.
CA Inc.
Research Staff Member, CA Labs
Intellectual Property and Standards
Council of Technical Excellence
W3C Advisory Committee Representative
Tele: + 1 603 823-7146
Fax:   + 1 603 823-7148
<mailto:kirk.wilson@ca.com>