[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

Re: XML Schema 1.0 and Relax NG Only Partially SupportContext-

• From: Rick Jelliffe <rjelliffe@allette.com.au>
• To: "Roger L. Costello" <costello@mitre.org>
• Date: Thu, 2 Jan 2014 11:03:33 +1100

I dont think the heading is a useful conclusion.  You usually only refer to a grammar by the class that parses all the members, not by a sub or superset.

Say you had a schema language, and it only allowed one element. It too would 'only partially support context free grammars'.  But that is like saying that the sound 'a' partially supports all the world's speech. The exceptions are more than the correspondences.

I have a new puppy: do i partially support all pets? Well, yes but really no: if i support only a small clear distinct well-recognised subset (ie 1) then it is misleading to say i partially support all pets. It muddies the waters that the classes are supposed to clarify.

Rick

PS I think there is also a useful distinction to be made between power and expressiveness: some "simple" constraints need great power, and some complex constraints need (a grammar type allowing) great expressiveness: regular expressions are not powerful but they are expressive. Xpaths are powerfull but not so expresive: an xpath that acted like a regular expression with axes and predicates would be best:  e.g.  a, child::b*, child::c

PPS Btw. Combining two threads, what about this: matching plus intersectability/openness?
<assert test='count(a) = count(b)  and not(a/preceding-sibling::b)'>
a^n,b^n  ignoring other elements
</assert>
It would be ok to say that schematron partially supports CFGs. Indeed, it may fully support them for all i know (eg if used with xslt2, and making use of function definitions).

• References:
  • XML Schema 1.0 and Relax NG Only Partially Support Context-FreeGrammars
    • From: "Costello, Roger L." <costello@mitre.org>