1 Interoperability: architectures and connections John Gilby, M25 Systems Team, LSE Ashley Sanders, Copac Team, MIMAS "Hyper Clumps, Mini Clumps and National.

1

Interoperability:architectures and connections

John Gilby, M25 Systems Team, LSE

Ashley Sanders, Copac Team, MIMAS

"Hyper Clumps, Mini Clumps and National Catalogues: resource discovery for the 21st century“

11th November 2004, British Library, London

2

Contents

• Overview of technical architecture of union catalogues (Copac & InforM25)

• Introduce Z39.50 to Z39.50 middleware & issues to consider

• CC-interop and JAFER– Installation, configuration & testing

• Results set issues and searching times

3

A reminder, Z39.50 is:

• a standard for information retrieval

• a client/server relationship– Z-client – stand-alone in PC or associated

with web server/user interface– Z-server - generally a module in library

systems

• a method for communication between disparate computer systems (such as a library catalogue and a user’s PC)

4

Copac• has 26 libraries (including large research, academic and BL, NLS)

• geographically covers whole of UK• JISC funded, administered by MIMAS• has “control” over indexes and searching process• can be searched via Z39.50• periodic data loads• live circulation data via Z39.50 = very successful

and popular with users• Copac V3 – experimental Z39.50 searching of

Copac and National Library of Wales

5

Incoming MARC records from contributing institutions

Record pre-processing:

standardisation & problem

identification

Copac database

Z-server, OpenURL& web interface

Formation of consolidated and

individual records & indexes

CURL database creation

MARC21 & UKMARC

Duplicate checks

pass/fail

web Z39.50

CURL/Copac database creation

6

Distributed catalogue• typically has up to 40 library catalogues (academic – CAIRNS,

InforM25, RIDING; Public - WiLL)

• regionally based

• funded by regional organisation

• rely on institutional catalogues for record standards, indexing and Z-server configurations

• some control over Z39.50 searching process

• data is as up to date as library OPAC

• ‘clump’ software combines result sets and presents them to user

• generally cannot accept queries outside of user interface

7

User

Copacsingle, large

database

Distributed catalogueZ-client software and

user interface

Z-server/institutional library systems

network

network

Union catalogues

8

Z to Z Middleware

Remote user

Z-client

Z39.50to

Z39.50Middleware

Institution Z-serverA

Institution Z-serverB

Z39.50

Z39.50

‘Local’ user web interface

e.g. M25 libraries

e.g. Copac V3

9

Connection Issues

• When to make connections ?• Which Z-servers ?

– selecting some/all, landscaping

• Access & Authentication– handled by middleware

• Timing of middleware response– user’s client is expecting single response– middleware has to wait for Z-servers to respond

before it responds to client– automatic time-out advisable

10

Search & Result Set Issues

• Query transformation– multiple Z-servers behave differently to an incoming

query– user sends query in their own ‘format’ (attribute set)– need to avoid failed searches– middleware transforms query to form suitable for

individual Z-servers

• Response aggregation– user’s client cannot know hits/Z-server– client must display origin of record– various options

11

and so to JAFER

• Middleware options for CC-interop:– graft Z39.50 server onto existing InforM25

software– develop completely new software– use existing available software

• JAFER Toolkit Project (JISC 5/99 Programme)– readily available & supported– could do most of what was required

12

Working with JAFER

• JAFER: http://www.jafer.org/– increased the JAFER logging facilities– established subsets of libraries for

searching– produced XSLT stylesheets

• Created new Copac Interface– copy of standard Copac web interface

tailored for testing JAFER

17

Search tests

• Search set 1 - Copac Z39.50 criteria– no query transformations

• Search set 2 - M25 ‘best practice’ settings– query transforms applied

18

Search test results — 1

• Access failed– variable: always, sometimes, occasional,

never– Talis & Aleph access problems – firewall problems

• Access succeeded– some searches received no response

19

Search test results — 2

• Response with Copac search settings– 203 searches carried out– 95 failed to return a result (0 or more

records)

• Response with InforM25 settings– 199 searches carried out– 3 failed to return a result (0 or more

records) 

20

Middleware benefits

• Simplifies access to range of catalogues• Query transformation improves search

success rate• Virtual catalogue staff can:

– provide centralised development and maintenance– identify and investigate problems– act as a central contact point

• Can interconnect the (JISC) Information Environment

• Potentially useful for a National Catalogue

21

Search problems/solutions• Users lose control of query• Search consistency

– failure of catalogues to respond– lowest common denominator or all options?– catalogues searching different fields– catalogues searching fields in different ways

• Standardisation– profiles eg. Bath Profile – work on index standardisation

22

Response times

• Improved access to resources– benefits end-user and library staffBUT– impacts on local catalogue– over-large result sets– duplication of material 

• Response times– impact on local catalogue searcher– impact on virtual catalogue searcher

23

Response time test

• Hourly search for ‘Austen’ – record time taken to obtain search result– does not include record collection or result processing

• Number of searches responding– c.90% within 2 seconds– c.4% within 4-27seconds

• Overall response time governed by slowest catalogue– Timeouts for slow to- or non-responding catalogues

24

Restricted searches

• Should all searches be sent to all catalogues?– control where searches are sent initially

– pre-defined search groups - by location/subject?

• Better to deal with large result sets through ranking and/or sorting?– which brings us back to response times…

25

Summary & what next ?

• JAFER tests - middleware works• Enables distributed catalogues to be

‘plugged into’ the IE• Dynamic resource selection is technically

feasible• Clump services interested• Further investigations:

– Response-time tests– Results processing

26

Further details

• Reports on the project website:http://ccinterop.cdlr.strath.ac.uk/documents.htm

• Copac Team:[email protected]

• M25 Systems Team:[email protected]