Re: PSVI

[Uche Ogbuji <uche.ogbuji@fourthought.com>]

Sean McGrath:

> We need clean layers darn it! The PSVI is a validation and
> a transformation rolled into one. With SGML we have
> a parse that was part validation, part transformation and
> part transclusion. Lets not go there with XML.
>
> Make non-validating XML 1.0 parsing the norm (It is anyhow, in my
> experience). Deprecate the doctype,  Make transformation &
> validation operations sit on top of the non-validating XML parse
> infoset and get 'em to specify the machinery with which
> to do it, *outside* the instance.

Sean put words to things not many would dare to say, but I think he's spot
on.

It seems one of the eikons of SGML-derived systems is mixing in a
convenient  bit of higher-layer processing into the parsing stage.  It
seems so much  the norm that before you know it something like XSchema
sneaks up to scare us off our curds and whey (no offence to those who did
a thankless job on Xschema).  I think this eikon needs be clast.

Parsing should be parsing and there should be a "canonical" infoset as a
result.  A higher layer can then come along and add that
attribute-defaulting people seem to think they need so desperately.
Another layer can come along and add type decorations.  Of course the
infoset would need a clean extension model, but I think this is quite
feasible.

If there were a clean and well-mapped pipeline of XML processing I think a
lot of the concern over intertwined yet out-of-sync specs and maybe even
PSVI infiltration could be satisfactorily resolved.


-- 
Uche Ogbuji                               Principal Consultant
uche.ogbuji@fourthought.com               +1 303 583 9900 x 101
Fourthought, Inc.                         http://Fourthought.com
4735 East Walnut St, Ste. C, Boulder, CO 80301-2537, USA
Software-engineering, knowledge-management, XML, CORBA, Linux, Python