Re: native encoding

----- Original Message -----
From: "Matthew Dovey" <matthew.dovey@las.ox.ac.uk>
Sent: Monday, March 04, 2002 12:24 PM


> Sometime ago I suggested that we should consider registering an OID for
> ONIX. The general response was that we already had an OID for XML and
> that as ONIX was an XML document we should use that OID. The particular
> schema could be specified using compSpec schema - rather than register
> an OID for any XML schema


>
> At the time this seems at logger-heads with the way that MARC was
> handled in that we have an OID for every MARC flavour out there.
> Strictly speaking ISO2709 (if I've got the number correct) is the record
> syntax whereas MARC21 vs UKMARC vs UNIMARC etc. are just "schemas" of
> the same syntax. The general consensus was that if we were starting
> again we'd probably distinguish between these.

<TripDownMemoryLane>
I will readily agree that the way we treat MARC and the way we treat XML is
inconsistent.  That's because the way we treat MARC stems from OSI days.
"Strictly speaking", ISO2709/Z39.2 is a TRANSFER syntax.  The definition of
tag semantics (an ISO2709 "implementation") is an abstract record syntax.
And strictly speaking, character set encoding should be handled by
presentation layer negotiation and transformation.

By the time we got to compSpec, we had pretty well abandoned OSI's formal
distinction between abstract and transfer syntax.  Out with the bathwater
went some of the useful parts of the presentation layer, such as character
set handling.  So XML, which is strictly a transfer syntax, was registered
as a record syntax (because that's the way we did it at the time).  And XML
schemas, which are strictly abstract record syntaxes, are new and different.
What's happened is that things moved up a level - what was transfer syntax
has become record syntax and what was abstract syntax has become abstract
schema.
</TripDownMemoryLane>

>
> However, this seems to be going the other way - we have three distinct
> concepts the record syntax/structure (MARC, XML, GRS.1), the record
> schema (MARC21/UKMARC, XML Schema, TagSet) and the character encoding -
> are we really going to suggest a distinct OID for every combination? Do
> we have enough OID's?

I didn't suggest that having separate OIDs for MARC21withMARC-8 and
MARC21withUTF8 was formally or abstractly the correct thing to do.  I
suggested that, since, as far as I know, the requirement to specify the
desired character set comes only from users of MARC21, separate OIDs was a
practical and efficient thing to do.

(I still think that if XML had been around - or fully appreciated - at the
time we wouldn't have done GRS.  I remember saying when GRS finally agreed
that "we have just reinvented SGML".)

j.

Received on Monday, 4 March 2002 13:53:17 UTC