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

Re: maps

• To: dareo@m... (Dare Obasanjo)
• Subject: Re: maps
• From: John Cowan <jcowan@r...>
• Date: Sun, 4 Aug 2002 22:49:01 -0400 (EDT)
• Cc: catcode@c... (J. David Eisenberg), simonstl@s... (Simon St.Laurent), xml-dev@l...
• In-reply-to: <8BD7226E07DDFF49AF5EF4030ACE0B7E06621EAD@r...> from "Dare Obasanjo" at Aug 04, 2002 01:04:04 PM

Dare Obasanjo scripsit:

> How does RELAX NG provide a mechanism for specifying operations on
> these datatypes? It would be easy enough to specify the lexical form
> of longitudes and latitudes then place them in a publically accessible
> XSD file but it would be useless since there would be no language for
> specifying value space relationships in the schema.

I think there is a misunderstanding here.

RNG schemas do not have a mechanism for defining datatypes, but rather
for using them.  To define a datatype, you must write code in a conventional
programming language that allows three questions to be answered:

1) Is string X in the lexical space of datatype Y?

2) Are strings X1 and X2 lexical representations of the same value of
   datatype Y?

3) Does datatype Y have the identity semantics of ID, IDREF, IDREFS,
   or none of the above?


By "datatype" here I mean "datatype with whatever constraining facets
the schema author specifies, out of the set allowed by the datatype
library author".  RNG datatypes may permit arbitrary facets, including
or excluding those known to XSD.

> From what I've seen RELAX NG supports datatype libraries and parameters
> to these data types but this seems to imply that each RELAX NG validator
> needs to know how to deal with type libraries so it isn't simply a case
> of writing a schema but instead actually altering the code for a RELAX
> NG validator for it to support a datatype library.

Jing and MSV, and probably any other validator written in Java, accept
datatype libraries that conform to the interfaces in the org.relaxng.datatype
package.  It would be straightforward to write a datatype library
for cartographic locations in Java.  There are analogous facilities for
Tenuto, the C# RELAX NG validator, and for COM-based validators (I don't
know if RELAXNGVB supports this or not).

> So interoperability
> goes to hell unless one can convince every RELAX NG validator to support
> your homegrown datatypes.

Interoperability of schemas that use datatype libraries depends on de
facto rather than de jure considerations, it's true (except for the
trivial built-in library that supports "string" and "token").

-- 
John Cowan                                   jcowan@r...
        "You need a change: try Canada"  "You need a change: try China"
                --fortune cookies opened by a couple that I know

• References:
  • RE: maps
    • From: "Dare Obasanjo" <dareo@m...>

• Prev by Date: Re: maps
• Next by Date: Re: Keeping ISO 8879 Alive (was RE: Markup perspective not code)
• Previous by thread: RE: maps
• Next by thread: Re: maps
• Index(es):
  • Date
  • Thread