RE: XDM serialization (was buried in microXML thread)

[David Lee <dlee@calldei.com>]

Thanks for reading and commenting Liam.
Yes function items are problematic and for a while I was so depressed I gave up on the dream of XDM Serialization.
But now I am more willing to compromise.   I think the value of supporting *many* XML types overweighs the inability to support *all*.
For function items I think a subset could be serialized such as   fn:concat#2   ... but much like user defined types which would only work if there was an in band or out of band schema,  arbitrary function items may require full XQuery lying around or worse, whatever proprietary language instantiated it.  And thats not even considering the in-scope (lambda) variable problem.

As for the delemiter strings,  yes I agree they could be nearly anything.  In fact I gave up arguing with myself what they need to be and 
just defered that decision to later ... because  they are largely arbitrary as long as they meet the constraints you give.
I still am on the fence as to predefining one or leaving them configurable.   Freedom vs 	responsibility !
I had not considered either content-length, chunking, or EXI
Food for thought ! Although I think any of those adds complexity (especially EXI ...)
Although I would be all for EXI being enhanced to support XDM ...
It just disappoints me greatly that XPath, XSLT, and XQuery have this great data model with no standardized way of representing it.
Thus everyone just deals with Documents.   Like that is all there is ...   XDM is wonderfully rich ... but all the XDM aware processors are stuck with proprietary only means of passing it around, which usually means it is not passed around.   Even something as simple as a sequence of documents, or <gasp horrid> a sequence of strings has no portable representation either in memory or serialized !
<rant rave rant rant rant rant >

Time for a beer


----------------------------------------
David A. Lee
dlee@calldei.com
http://www.xmlsh.org


> -----Original Message-----
> From: Liam R E Quin [mailto:liam@w3.org]
> Sent: Monday, July 16, 2012 5:44 PM
> To: David Lee
> Cc: Uche Ogbuji; xml-dev@lists.xml.org
> Subject: XDM serialization (was buried in microXML thread)
> 
> On Sun, 2012-07-15 at 14:42 +0000, David Lee wrote:
> 
> > See my cross posted reference to
> > http://xml.calldei.com/XDMSerialize
> 
> This also misses function items and other XDM 3 additions.
> 
> > I had an "Ah Ha" Moment last week when I realized that the UTF8 BOM
> could serve as such a separator.
> 
> You could use any string if you declared it, ala MIME conventions, and
> had a way to escape it, and a way to escape the escape.  SGML and XML
> get that sadly wrong, and even microxml, as an XML subset, can't grok
> &#ddd; used as part of an element name.
> 
> You need a string that cannot occur in the middle of a document (out of
> band) or one that can be escaped (in band, future-proof but more work)
> or content-length (hard to stream) or a chunked encoding (pain to decode
> but I think more promising). Maybe EXI could be made to work?
> 
> Liam
> 
> --
> Liam Quin - XML Activity Lead, W3C, http://www.w3.org/People/Quin/
> Pictures from old books: http://fromoldbooks.org/
>