[standards-jig] schemas and sequence

Bob Wyman bob at wyman.us
Mon Feb 17 21:04:59 UTC 2003


> The question immediately arises: are these 
> schemas to be construed as normative or descriptive? 

This question of choice vs sequence is precisely why it will be useful
to wean Jabber and other similar efforts off of DTDs and convert to XML
Schema. The fact that DTDs don't express these important syntatical
details is precisely why DTD's and SGML were rejected way back in the
early/mid 80's when they were proposed as an alternative to ASN.1 for
defining message content for the ill-fated X.400 email standards. It has
been terribly frustrating to see DTD's raise their heads again over the
last few years and re-introduce all the same problems that we had with
them 20 years ago! Today, we may be stuck with the horrendous
inefficiency of textual formats derived from SGML, but, at least XML
Schema gives us the ability to remove the ambiguities that resulted from
the use of DTDs. XML may still be fat, but XML Schemas means that it
doesn't have to be sloppy.

The reality is that since you've got code in the field, you are probably
stuck having to say that sequence is "descriptive" not "normative" when
describing things that have already been implemented widely. However,
every effort should be made to ensure that any new standards, or new
versions of old specifications carefully consider whether sequence
should be required. Often, it is useful to require sequence since
processors, if they can rely on sequence being enforced, are able to use
that reliance to improve both the correctness and performance of their
code.

		bob wyman





More information about the Standards mailing list