ZIG Meeting
12/8/2000
Library of Congress
Explain
Chair: Bob Waldstein
Notes provided primarily by Les Wibberley.
Overview - Ray
- manifestations
- 1988 version - explain was referred to as a future service
(singular)
- 1992 - future services (plural) mentioned
- 1995 - separate facility actually defined
- 1995 - explain facility
- explain database
- accessed via search and retrieval
- attribute set and syntax
- ISO SR roots of this in the form of Help (tasteless joke deleted)
- Rationale
- Denis lynch paper in 1997
- described concepts
- lots of complexity in explain thought necessary several years ago
- complexities
- access info , termlist associated DBs, sub DBs, per element details,
processing details, etc.
- issues
- model
- complexities
- what information does a client need
"Bob's Perspective"
Bob Waldstein
- personal experience
- hundreds of databases with diverse design
- explain met the needs to be able to search all these databases
- use mostly attributes and element sets
- implemented other features
- have had a wide range of requests from outside the zig
- others are finding out databases using explain
- Explain is not really part of the protocol
- search and retrieval implementation
- much of it only explained in Denis's paper
- There is a core set of functions used
- Bob beginning to give up on it
- has had it up for about 7 years, but know of no one else interoperating
with it
- will continue to use for internal use
- not sure about need for interoperability
Explain-Lite
Poul Henrik Jørgensen
- background
- ONe project charged to implement explain
- requires significant effort to implement at both client and server
- maintaining the explain database requires tools
- very few client request the explain data
- most clients use proprietary files from target
- lots of effort to implement, and not used much
- objectives
- define common extensible config. file for z clients
- much simpler to implement and maintain
- usable for both z clients and humans
- minimal impact on z profile
- solution
- xml structure with relevant config. information
- delivered in init response packet
- also presented via web servers with xsl style sheets
- specification:
- http://www.bibsys/z/explainlite.dtd
... will be put online with slides
- xml example
- xsl display resulting from style sheet
- summary
- explain lite developed as alternative to proprietary client config. files
- same info may be used as user info via the web
- should be aligned with emerging generals standards such as UDDI and WSDL
(Web Services Description Language)
UDDI
Matthew Dovey
- Universal description, discovery, and integration
- motivation from the e-business community
- consortium of major computer system vendors
- not a new concept or technically rich
- lot of enthusiasm and support, which is more important for interoperability
standard
- e-marketplace model
- businesses advertise services
- customers/clients locate services
- client software obtains protocol level information needed to connect
and interact with the service (like explain)
- goal is that this happens automatically behind the scenes
- technical
- data are in xml model
- api specified via SOAP
- uddi over xml over SOAP over http over tcp/ip
- three reference implementations
- T-models specify a template for the contact information
- link to Tmodels is a windows-like link
- explain and uddi
- alternative mechanism for publishing explain and lite
- uddi frontend gateway to existing explain
- why? increased visibility of services
- asn.1 and WSDL
- ASN could be recast in WSDL
- could produce the core spec better suited to xml encodings
- next generation tools would build soap/xp layers automatically on discovering
the protocol from the uddi register
- raises the visibility of z39.50
- Matthew will be experimenting with this
- directories
- z39.50 service registry?
- ill/ncip registry?
- perhaps as frontend to explain/ldap
- technical issues
- very poor search interface at this time - maybe use z39.50.
- metadata model is very naive
- issues of global distributed registry (replication)
- related t-Model structure which is not very rich yet ( reflecting
the relationship between protocol, attribute sets, profiles, and service
specifics)
- But lots of interest and backing. This is where to be seen to attract
customers
- questions
- the api is used to interact with the registry and the information coming
back are specified, the registry itself is not specified
- how to know where to start to find this
- there is a whole replication of uddi registries
- discussing interaction between this and ldap
- who is working on this? plan to bring the specs to w3c, but for
now group of major computer vendors (many major players)
- lots of companies already registering on the registry
- all information registered is public domain
- open architecture to add information
- repositories replicated or referred?
discussion
- explain issues
- do we need an interoperable explain?
- what parts of explain are needed? for what?
- example: drop support for things like negotiating extended services
- is the search and retrieval model correct?
- should it be part of the z39.50 protocol?
- Ralph: tried to do explain, and it was way too much work
- Lennie: agreed with Ralph's position
- But most clients are not flexible enough to dynamically react to discovering
this on the fly within the session
- Mike and Joe agreed
- Joe indicated that they were now using Explain Lite, but a different
flavor of it
- Is search/retrieve the right model for Explain?
- sending stuff in init for explain lite.
- but these various versions use different formats
- explain lite also gloms all the different databases together
- Explain doesn't solve the whole discovery problem
- perhaps consider use of UDDI as an alternative
- use of z39.50 search and retrieval model implies a boot-strap problem
- UDDI and Explain are two different ways of looking at the same information
- Mike: proposal for adding something new to Explain, from the Thesaurus
project
- Ralph: tried to implement the full explain, and failed
- perhaps focus on identifying what is necessary and complete, and grow
experience
- Ray: Explain is a more controlled service, client has some control
over what is returned, e.g., after certain date, for specified database; explain
lite doesn't scale up well for large number of databases
- perhaps pare down existing explain and add explain-lite
- keep both of these as alternative approaches to explain
- Sebastian - perhaps cast explain lite as a new thing, pared down from
explain
- consider explain-lite as the executive summary aspect of explain; augments
it
- Bob: Alan Kent had a bunch of more new categories he was using
- there has never been any provision for getting explain intact without
using explain syntax
- Sebastian: should be able to define an XML equivalent pretty easily
- extensibility of explain (Mike)
- Ray: there is the capability to add new categories via OID
- Bob: wrapup
- put out proposal on the list to get a single explain-lite definition,
add new categories, support XML syntax
- Ray: propose new version of explain, with sections deleted, defined by
explain-2 record syntax, create simpler subset
- Ralph: endorses profile view, rather than cutting sections out.
This is less disruptive. Just focuses on the key things really needed
- Ray; make this a self-contained profile and not point to the original
explain standard
- Bob will lead this effort