RE: [w3c sml] [4688] Consider using xml:base

All (except Sandy :))

Please speak up now if you disagree with Sandy's proposal. You may want to read my comments on Sandy's proposal.

Thanks!
Pratul

From: Sandy Gao [mailto:sandygao@ca.ibm.com]
Sent: Friday, September 07, 2007 8:58 AM
To: Pratul Dublish
Cc: public-sml@w3.org
Subject: Re: [w3c sml] [4688] Consider using xml:base


Pratul and all,

> 1.       <baseURI> allows attribute extensions, and this will be
> lost if xml:base is used

Yes, and the rest of the XML world has lived with that for many years. (xml: id, lang, base, ... are all attributes)
[Pratul] Members should consider what is best for SML.  We should avoid gratuitous divergence from existing standards but we should feel free to define new things if required.


> 2.       The xml:base spec uses the definition of URI from RFC 2396
> but the definition of xs:anyURI in XML Schema 1.0 uses RFC 2396 as
> amended by RFC 2732. So there is a possibility (I don't know for
> sure since I haven't read 2396) that xml:base is not aligned with xs:anyURI

There's some weirdness in the xml:base spec. It only mentions 2396 in most cases, but also mentions 2732 in one place.

Yes, it's possible to have such inconsistency, but this would happen between any 2 specifications we reference. For example, if XPath 1.0 and Schema 1.1 are used together; ...

(This one feels even weaker, because 2732 really isn't much different from 2396.)
[Pratul] I found that XML Base has issued an Errata (http://www.w3.org/2001/06/xmlbase-errata) that switches to the definition of URI in RFC3986. This is consistent with SML and SML IF specs - both of which reference RFC 3986.  We already have the possibility of inconsistency with XML Schema 1.0 since it uses RFC 2396 as amended by RFC 2732.


> 3.       xml:base does not support IRIs, so it can't be used if we
> decide to support IRIs in SML IF ( bug 4632  )

It does support IRIs in a way similar to anyURI. xml:base allows non-ASCII characters; its values, after escaping, become valid URIs.
[Pratul] My point was that xml:base does not follow the IRI RFC (3987).



So I don't see any of the above as significant problem. On the other hand, reinventing an existing concept, which is widely supported (including infoset), seems more problematic to me.

So counter-proposal:
- Remove smlif:baseURI
- Add a reference to xml:base in "modelType" (to allow <model xml:base="...">)
- Use infoset [base URI] property in places where "smlif:baseURI" is currently used.

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


public-sml-request@w3.org wrote on 2007-09-07 12:28:00 AM:

> All
> This is an attempt to instigate consensus on the resolution of this
> bug. We don't have any specific proposal for this bug, so I am going
> to make a proposal based on my comments in the bug. The proposal is
> that SML IF should not use xml:base and continue to use the existing
> <baseURI> element because
>
> 1.       <baseURI> allows attribute extensions, and this will be
> lost if xml:base is used
> 2.       The xml:base spec uses the definition of URI from RFC 2396
> but the definition of xs:anyURI in XML Schema 1.0 uses RFC 2396 as
> amended by RFC 2732. So there is a possibility (I don't know for
> sure since I haven't read 2396) that xml:base is not aligned with xs:anyURI
> 3.       xml:base does not support IRIs, so it can't be used if we
> decide to support IRIs in SML IF ( bug 4632  )
>
>
> Please speak up now if you disagree with this proposal.
>
> Thanks!
> Pratul