Element sets, schemas, and record syntaxes

Hi,

I was just looking through the Z39.50-2003 spec, noticing a few changes
here and there, such as the introduction of a URI as an alternative way
to identify the schema to use (see CompSpec).

I then looked at Explain, and noticed the schema there can still only
be an OID.

Q1: Theoretically, should the places where a schema OID is present in
Explain (for example, the non-key 'schema' field of ElementSetDetails)
really now be OID or URI (a CHOICE)?

I often read the Explain ASN.1 as a way of understanding the Z39.50 data
model, as it defines in effect an E-R diagram of the Z39.50 concepts
(including what should be keys). Reading ElementSetDetails again, I
notice that the following is a key: databaseName, elementSetName,
recordSyntax. The schema OID does not form part of the key. This means
that an element set name for a specific record syntax must always use
the same schema.

Note that in a present request you can specify the schema OID in CompSpec,
the preferred record syntax, the element set name, and the databasename
(the database name was in the search that formed the result set).
Reading things like the holdings schema and with XML in general, I suspect
the world has moved a bit.

Q2: Can you specify the "XML" record syntax, the same element set name, but
request a different schema to get back different XML content?
Or does the element set name and record syntax (eg: "XML") uniquely
identify the schema to use?

I wonder if it is now more accurate to say the schema is now part of the
key. If not, there does not seem much point in being able to request a
particular schema.

If the schema is now part of the key of ElementSetDetails in Explain, its
backwards compatible (all old implementations are still correct), but the
'schema' of ElementSetDetails should now be a part of the key (and possibly
now be allowed to be a URI, not just an OID).

Note, the same sort of arguments for XML are also valid for GRS-1. There
is logic in asking for the 'full' or 'brief' version of a record sent as
GRS-1 accoring to different schemas.

Thanks!
Alan

ps: I am not proposing to change the Explain record definition to introduce
URIs - I just want to understand the correct model.

Received on Tuesday, 17 February 2004 21:29:44 UTC