Re: [xml-dev] Need a language whiz: An XML Schema "specifies" how data is to be structured? "describes"? "constrains"?

On 08/01/2018 11:14, Tony Graham wrote:
> On 08/01/2018 09:20, Rick Jelliffe wrote:
>> I still don't see what the novelty of "epischemas" is supposed to be,
>>  apart from giving a name (which is welcome).   James Clark did this
>>  parallel grammar technique for HTML IIRC, as a way to reconstruct
>> SGML's

I don’t want to boast about it, but I consider naming this thing quite a 
feat. As the jokes go, naming things is one of the hardest problems in 
computer science after all!

> 
> IIRC, Gerrit did refer to James Clark's precedent (and its absence of a
> name) in his talk, though I can't find a mention of it in either of
> Gerrit's slides or paper.

We’d have to watch the full, excruciating video in order to find that 
out. I don’t remember either.

What I did mention in the paper is that epischemas might resemble 
architectural forms (whose name, btw, is quite pretentious, given their 
relatively minor merit in computer science ;-) superficially, but 
actually epischemas (local constraints for elements with the same name) 
are the inverse of architectural forms (different element names reduced 
to a common grammar).

> 
>> exclusion exceptions (i.e. that an <a> could not contain an <a> etc)
>>  without combinatorial explosion.  Schematron supported multiple
>> patterns in parallel from the start.   DSDL's NVRL was based on
>> selecting different sections of documents and running them through
>> different schemas (including in parallel).  Lloyds of London
>> Financial Markets use of layers of increasing complexity (IIRC a
>> basic grammar, then a complex grammar, then a Schematron) to weed out
>> bad transactions efficiently was widely reported.  And the idea of a
>> Bloom filter is basic CS knowledge.
> 
> Everybody here is pro-Schematron, but Gerrit's stated reason for using
> multiple schema bundled with NVDL is context-aware content completion in
> editors.

The other reason was that some problems, such as modeling the permitted 
TEI div types after DocBook’s part/chaper/section elements, are more 
naturally specified as a grammar than as Schematron rules. Not all XML 
grammar languages qualify for modeling it though. You need an anything 
pattern that allows exclusions. These are available in Relax NG and XSD 1.1.

I’m a pro-Schematron guy, btw.

Gerrit