W3C home > Mailing lists > Public > www-webdav-dasl@w3.org > January to March 1999

Why Z39.50 will not work for DASL

From: Jim Davis <jdavis@parc.xerox.com>
Date: Fri, 29 Jan 1999 17:49:59 PST
Message-Id: <3.0.3.32.19990129174959.0097b780@mailback.parc.xerox.com>
To: www-webdav-dasl@w3.org
In earlier email, I had raised the possibility that Z39.50 might suffice
for all of DASL's requirements.  After further consideration, I have come
to the conclusion that while Z39.50 might be able to meet the current
requirements, it could NOT meet some of the likely future requirements.
Therefore it makes sense to continue to develop the DASL protocol.

The rest of this message will likely only make sense to those with a strong
background in Z39.50

My objection to Z39.50 is based on the inflexiblity of the Type-1 query.
In this query, one may provide up to six 'attributes':
  * use (which is similar to a DAV property)
  * structure (which is somewhat like a datatype)
  * relation  which is like a DASL comparison operator)
  * truncation 
  * position 
  * completeness

Thus one might search the "author" use attribute for a word (structure)
that is equal to (relation) a search term that is right truncated, at the
beginning of the field, and need not occupy the entire field (completeness).

My objection is that I don't think this six-attribute format will
generalize well to allow search of structured properties, e.g. to find an
string within (and only within) the locktype element of a lockdiscovery
property.  While such structured property queries are not in the
requirements for the current DASL, it is clear that they will be, someday.
Committing to Z39.50 now would make it hard to meet this future
requirement.  It would be a dead-end.

One could of course define a new type of query in Z39.50, but I think this
would be hard, for both technical and institutional reasons. It's not clear
any of the existing Z39.50 vendors would buy into this extension, and then
we'd have the same bad consequence I was hoping to avoid, namely
fragmentation of the searchable internet resources.  Instead, we can avoid
this fragmentation by building a gateway from DASL to Z39.50.  Thus,
existing Z39.50 servers will be searchable by DASL, and hence the DASL
protocol will suffice to search these online resources.  (A gateway from
DASL to LDAP might also be useful.)

Best regards

Jim Davis




------------------------------------
http://www.parc.xerox.com/jdavis/
650-812-4301
Received on Friday, 29 January 1999 20:50:18 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Sunday, 22 March 2009 03:38:04 GMT