MARC Formats Interest Group
Saturday, January 10, 2:00-4:00 p.m.

The main discussion topic was the migration of data from one library automation system to another. The issues were addressed by a panel consisting of Janet Fox (University of Chicago), representing a library that has planned and executed data migration to a new system; Barbara Tillett (Library of Congress), presenting the perspective from the national library which is in the early stages of planning a data migration; Joe Altimus (Research Libraries Group), representing a bibliographic utility's perspective; and Robert Pillow (VTLS) who presented a vendor's perspective.

Barbara Tillett spoke first, as she had to be at another meeting. She discussed the general aspects of LC's search for a new ILS, and gave examples of their current record structures (which are not all standard USMARC). She described the local fields that they will need to carry over, as well as the manual data (shelflist, primarily) that still remains. Their shelflist includes most of their holdings information, including call number and location, and they estimate it will take 10 years to migrate this data. Their target date is Oct. of 1999 to have their new system up and running (the old ones cannot last beyond 12/31/99) and clearly they are quite anxious about this timetable.

Janet Fox of the U. of Chicago presented the most interesting report. They went from a 20-year-old locally created system to Ameritech's Horizon system in 1995, one of the early implementors. Acquisitions is not yet up for them, but they have completed all the data conversion. (Acq data is being carried in local bib fields for now). Their previous system used USMARC for most cataloging, except authorities, but other data was in non-standard fields. When they migrated to Horizon, they built their copy records from their old system records, but imported a new authority file from BNA. They're now cleaning up duplicate authorities because of multiple records for names and subjects.

Because they had a real systems office, as well as locally developed output programs for their old system (designed to export to OCLC), they were able to use these to migrate their data to Horizon. In their first phase, they moved most of their local data to bibliographic 9XX fields (and an "enhanced" 852, used to create copy holdings records), then in subsequent passes have moved the 9XX data to other records. The same strategy was used for circulation data, which went from bib 999 fields to item records (the 999 was then deleted).

One interesting wrinkle was that they had full support for USMARC holdings 85X/86X pairs in their old local system (though some holdings was carried in 866/7/8). These fields, with local extensions, drove their serials control system. Because Horizon does not support full USMARC holdings, and only allowed retention of *current* pattern information, all the retrospective information that they'd built over time for prediction of issues had to be deleted. This limitation of Horizon has created some difficult display issues for them as well.

Joe Altimus of RLG outlined some ideal migration steps for RLG libraries moving to new systems:

  1. Notify RLG that the process of looking for a new system has begun
  2. Include an RLIN-aware person on search groups, and determine whether RLIN's requirements have changed
  3. Include in specifications the exchange of USMARC data and preservation of RLIN-MARC extensions
  4. Test USMARC exchange, including extensions
  5. Selection of a system should include contract agreements to maintain USMARC conversion with RLIN extensions
  6. Notify RLG of schedule for conversion
  7. Consult RLG about making the transition
Some common problems encountered when systems change:

Robert Pillow from VTLS gave a vendor's perspective on data migration. His first point was that the most critical element is not the vendor, but the library staff. Their knowledge of their old system and data will determine the outcome. He stressed the importance of well-defined and realistic expectations of outcome, and that staff need to have some political clout internally to make decisions and get support for them. He described working with a large European library, where three non-MARC systems were converted into one MARC compliant system, in an environment where they could not even create a good test database.

Question period tidbit: Kay Guiles, in response to a question, reported that LC will be sending their 3.8 million record Pre-MARC file to OCLC for a "batchmatch" type project. They expect 1.2 million records back. These will overlay records in the Pre-MARC file, and will be available for purchase through CDS and probably on the utilities. When asked when the LCCN changes will come about, he said probably by 1/1/2000, they hope (depending on what their chosen vendor can handle for control numbers).

Unofficial report by D. Hillmann