The Wellcome Library recently activated a system for the online ordering of its archive and manuscript materials.  Colleagues may be interested in our experience setting up and operating the system so a summary is appended below.
Do contact me if any further information is needed.
Chris Hilton

Dr. Christopher Hilton 
Senior Archivist, Department of Archives and Manuscripts 
Wellcome Library for the History & Understanding of Medicine 
The Wellcome Trust 
210 Euston Road 
LONDON NW1 2BE 
Tel.: (+44) 020 7611 8481 
The Wellcome Trust is a registered charity, no. 210183. 
Its sole Trustee is The Wellcome Trust Limited, a company 
registered in England, no. 2711000, whose registered office 
is 215 Euston Road, London NW1 2BE. 

At the Wellcome Library, the Archives and Manuscripts department operates an ISAD(G)-structured archival database side-by-side with a large MARC-based database of printed material.  Online ordering of printed material went live in summer 2003.  As the next stage of the project it was decided to extend online ordering to archive material; and, if possible, to devise a way in which the two systems could co-operate and readers order archive material using the same interface as that for printed material.  The system described below went live in November and after a few months' successful trial we are in a position to publicise it more widely.

The bases of the system are XML export from the archive database and XML harvesting into the MARC database.  (The specific products used are CALM for the archive database and Innovative Interfaces Inc.'s Millennium catalogue and XML Harvester product on the printed materials side; the principles should, however, be applicable to other platforms.)

A hitlist was produced comprising all Item level records in the archives database (this level being what the Library uses to indicate a producible unit - a volume or file, for example) and these were exported in XML format to an OAI repository on the archives database server.  In this format all fields in a record are exported, tagged simply with their database field names.

Scripts transform the archive data according to a local mapping of XML to OIA MARC21 tags.  From here, records are harvested using XML Harvester and loaded to the catalogue of printed materials.  This creates a corresponding record in the MARC database for every requestable item in the archive system.  The MARC21 record does not comprise the complete ISAD(G) record: only a small selection of fields needed to print out the order slip are harvested so that stack staff can identify the material easily (reference, title, date, box number etc.).  Readers can then place online requests for the item.  The record of their requests is maintained in one place rather than across two systems and provides a single set of statistics.  We have initially set a limit of twenty items which can be requested at one time.

The two corresponding records in the different systems are linked using the archive database Reference field (this being policed by the system to ensure that it is unique.).  On the record display in the archives database's web interface, Item level records have a clickable link, instructing readers to click here to request the item.  This masks a script that instructs the browser to take the contents of the Reference field, go to the database of printed materials and carry out a search for that reference string.  Similarly, the MARC21 record when viewed through the web interface contains a link back to the archives database constructed in the same way: a script instructing the browser to take the text from the Reference Number field and search for it in the archive database.

The ultimate test of the system, of course, is how well it works in practice and we have found it to work well.  Readers can request material via our web interface at a time that suits them (particularly valuable when visiting London for a short period, when it is advantageous to have items produced in advance) and use a system that is already familiar from its use with the collection of printed material.  Stack staff are able to say goodbye to the handwritten slip, which inevitably is sometimes faint or illegible.  The system also bounces back orders for things already in the reading room, eliminating many useless journeys for stack staff.  Records of what has been ordered are kept in one central location, facilitating better stock control and statistics gathering.  Colleagues working in similar environments to the Wellcome Library, where co-operation with an existing library database is indicated rather than use of an archive system's own ordering module, may well find this an approach worth exploring.