Re: [cgmo-webcgm] ISSUE: must conforming XCFs be validatable?

[Benoit Bezaire <benoit@itedo.com>]

Hi Lofton,

I thought about this some more and what you are in fact proposing is
that a conforming XCF file must not only be "well-formed"
(http://www.w3.org/TR/REC-xml/#sec-well-formed) but also a "valid XML
document" (http://www.w3.org/TR/REC-xml/#sec-prolog-dtd ).

From an authoring tool perspective, this may be difficult to achieve.
It is easy to achieve when there is no metadata, but more difficult
when there is metadata.

I doubt authoring tool will always be in a position to extend the
DTD so that the XML namespace metadata passes DTD validation. I think
this could cause more damage than good when exchanging XCF files.

What is an authoring tool suppose to do when he loads in a file with
this at the top?:

<?xml version="1.0" standalone="no"?>
<!DOCTYPE ...
[
<!ENTITY % grobjectExt "| customNS:customElement" >
<!ATTLIST grobject
  xmlns:customNS CDATA #FIXED "http://www.example.org/customNS";
  customNS:customAttr CDATA #IMPLIED >
<!ELEMENT customNS:customElement EMPTY>
<!ATTLIST customNS:customElement
  xmlns:customNS CDATA #FIXED "http://www.example.org/customNS";
  info CDATA #IMPLIED>
 ]>

Maintaining these DTD extension sounds like a lot of work on authoring
tools for the only purpose of XCF versioning... I know you are very
sensitive regarding feature creep, but this is a big one! This
proposal is in fact a lot more work on vendors than most of the other
proposals recently sent to the group.

Why not simply say that: a conforming XCF file must specify its
version either by explicitly specifying the 'version' attribute or by
including a DOCTYPE that references a webcgm DTD.

Thoughts

-- 
 Benoit   mailto:benoit@itedo.com


Thursday, June 16, 2005, 7:54:20 PM, Lofton wrote:

LH> [...oops, I goofed last time and sent it to list, instead of
LH> Ben ... noproblem, I think the choices are clear enough and ready
LH> for issuediscussion and resolution by the TC...]

LH> WebCGM TC,

LH> We resolved in 20050615 telecon that every conforming XCF
LH> must identifyits version.  That raised the question of "how", and
LH> theproposal for the original issue was:  MUST contain DOCTYPE
LH> (w/internal DTD or external ref. to DTD) and SHOULD also use
LH> 'version'attribute on <webcgm>.  This "how" issue hides amore
LH> fundamental issue:  must every conforming XCF contain DOCTYPEthat
LH> defines or references its DTD, i.e., be validatable.  

LH> The answer will mostly determine the answer to "how" every
LH> XCFidentifies its version.  So we agreed that I would circulate
LH> the newissue, split out from the original issue.

LH> ISSUE:  must every conforming XCF contain DOCTYPE that defines or
LH> references its DTD?

LH> ALTERNATIVES:
LH> Alt.1:  YES.
LH> Alt.2:  NO.

LH> Recommendation:  Alt.1, YES.

LH> Regards,
LH> -Lofton.

LH> ----- original issue, resolved "must identify version"-----

LH> At 04:40 PM 6/9/2005 -0600, Lofton Henderson wrote:
LH> Source:  Editors.

LH> ISSUE:  must every XCF instance identify its version?

LH> In ch.4, the examples 4.1, 4.2, 4.3 show <webcgm> tags
LH> without'version' attributes.  The examples also start and end with
LH> the<webcgm> and </webcgm> tags  (except 4.2 with a piece
LH> ofinternal DTD subset defining an extension element, which btw
LH> looks to melike wrong syntax.)  The examples do not show the XML
LH> declaration,and do not show a DOCTYPE (which presumably would
LH> contain an externalreference to the WebCGM XCF DTD.)  [Side
LH> suggestion:  make theexamples be complete xml files.]

LH> In 4.2, the DTD defines:  version CDATA #fixed 2.0

LH> So if the XCF instance contained a DOCTYPE referencing the
LH> DTD, theversion would be defined.  (#fixed means:  if 'version'
LH> ispresent, it must be 2.0; if not present, 2.0 is the default).

LH> If an XCF instance contains neither a DOCTYPE nor 'version'
LH> attribute on<webcgm>, then its version is unknown.

LH> Question.  Is WebCGM 2.0 going to allow such un-versioned XCFinstances to be conformant?

LH> Alt.1:  no, fix the current ch.4 text.
LH> Alt.2:  yes

LH> RECOMMENDATION:  Alt.1, no -- every conforming XCF must somehowidentify its version.

LH> If Alt.1 is accepted, then "how"?  There are several waysto
LH> achieve it, here's one proposal...

LH> PROPOSAL.  1.) Require that every conforming XCF must contain
LH> aDOCTYPE referencing its DTD (and may also have internal DTD
LH> subset forextensions, of course).  And, 2.) Recommend that the
LH> 'version'attribute be used on the root element (<webcgm>).  I know
LH> that#2 is redundant with #1, but there are circumstances and
LH> scenarios whereit is convenient to have the information there in
LH> the XCF instance,instead of in some remote referenced resource
LH> (the DTD).

LH> Regards,
LH> -Lofton.