RE: Open systems / Freedom ( was RE: The Web as an Application)

 

> -----Original Message-----
> From: David Carlisle [mailto:davidc@nag.co.uk] 
> Sent: September 23, 2013 10:34
> To: Rushforth, Peter
> Cc: public-xmlhypermedia@w3.org
> Subject: Re: Open systems / Freedom ( was RE: The Web as an 
> Application)
> 
> On 23/09/2013 14:22, Rushforth, Peter wrote:
> > I realize that XML can not be said to have a "schema", in the XML 
> > Schema, DTD, RelaxNG sense of the word.
> 
> As I say, you should use a different word if you want to 
> avoid confusion:-)

Not sure what the appropriate term might be.  Semantics is very heavily overloaded.
Grammar doesn't seem to completely fit.  "Syntax and semantics" might do.
XML, like other media types on the web, is defined by its media type registration:
http://www.w3.org/2001/tag/doc/mime-respect-20060412#media-type

Do you have a suggestion?

> 
> 
> > If the designers of XML had intended that the xml: 
> namespace was a "no 
> > fly zone", XML Processors would reject content in that 
> namespace which 
> > they do not understand.
> 
> I don't accept that argument at all: At that level you are 
> just talking about the parser and there are no "understood" 
> attributes at that level so long as it is well formed

xml:lang has a defined meaning.  Not all applications of xml use it,
but if they do, its meaning is set by the XML specification.
The parser "understands" it enough to let it pass without comment.

> <a zzz="sss" xml:zz=".." zz:ww="" xmlns:zz=".."/>
> 
> will be accepted by the XML (Namespace) Processor, the fact 
> that xml:zz is parsed without error tells you nothing more 
> than the fact that zz:ww="" is parsed without error. The XML 
> Namespace is predeclared but that is the only difference, 

No. As mentioned above, @xml:* are predeclared and predefined.

> There are many applications where using any colon-prefixed 
> attribute is a pain (anything related to html for example) so 
> for those it is much simpler to use href="..." than 
> ???:href="...". Especially for applications of XML on the 
> web, having a relationship to HTML doesn't seem that unlikely.

Absolutely.  But since (you mentioned some time ago)
web browsers already include full XML parsers, they might come
to be able to recognize links in XML as slightly different than
in html.  XML-based media types would have XML links, and html
would have its own defined links.  

So, a relationship by reference would be possible and fruitful, I think.


> 
> Applications that are already handling namespace aware XML 
> could probably cope with href="..." as well (as shown by 
> xhtml or mathml or many other xml vocabularies that include 
> link semantics)
> 
> If you want to use a namespace prefixed version the choice between
> 
> xml:href=".."
> 
> xlink:href=".." xmlns:xmlink="..."
> 
> newns:href=".." xmlns:newns="..."
> 
> The first is a minor, almost trivial, syntactic 
> simplification for a namespace aware system, with a price of 
> all sorts of backward compatibility and political 
> coordination issues. I can't seriously consider that as a 
> viable option.

You and lots of others, I guess.  But for the record, what are the
backwards compatibility issues you see?   It might be wise
to put those on the wiki.

> 
> The second is done, so xlink is there if you want it, 
> presumably you want the semantics for this to be a bit 
> different, which leaves the third option.

IMHO the third option will be no more likely to succeed than xlink.
XLink after all is simply missing some attributes.  Probably has a few
too many attributes as well.  

IMHO that may leave this option:

Defined attributes in no namespace and a new media type.

The latter is not my preferred route, because it doesn't really help XML, per se.

But, if there is interest in your option 3, I don't want to be the block on that.
If the group thinks that's the right direction, please move on with that.

Peter

Received on Tuesday, 24 September 2013 14:07:26 UTC