Adding a Z39.50 target to Koha for copy cataloging

Z39.50 targets are added in Administration → Additional parameters → Z39.50/SRU servers. A target needs five things to work: hostname, port, database name, record syntax and character encoding. Get the database name or the syntax wrong and searches return nothing at all — with no error, which is why most broken targets look like network problems.

Go to Administration → Z39.50/SRU servers → New Z39.50 server and fill in the following. These values are the ones Koha ships in its own optional sample data.

The same host serves authority files on different databases: NAF for names and SAF for subjects, both with record type Authority. Add them as separate rows — a target is one database, not one server.

1. 1Go to Cataloging → New from Z39.50/SRU.
2. 2Tick the target you just added in the list on the right.
3. 3Search by ISBN first — it is the least ambiguous field and the fastest way to prove the connection.
4. 4Click Import on a result to open it in the MARC editor, then edit and save it as your own record.

Set Server type to SRU and the target speaks HTTP instead of the binary Z39.50 protocol. That matters when a firewall allows web traffic but not port 210, which is common on a university network.

An SRU target additionally needs SRU search fields mapping — Koha's internal field names mapped to the indexes that target actually publishes. Without it, every search is sent against an index the server does not have and returns zero results. Koha's sample SRU entry for the Library of Congress uses:

title=dc.title,isbn=bath.isbn,srchany=cql.anywhere,author=dc.author,issn=bath.issn,subject=dc.subject,stdid=bath.standardIdentifier

Every SRU server publishes its own index list in its explain response, and they differ. Copying another server's mapping is the most common reason a correctly configured SRU target finds nothing.

Z39.50 has almost no error reporting a user ever sees. A wrong database name, an unsupported syntax and a blocked port all produce the same screen: a search that completes with no results. So diagnose by elimination rather than by reading the message, because there is no message.

Test the connection from the Koha server directly when you want a definite answer:

$ nc -vz lx2.loc.gov 210 Connection to lx2.loc.gov 210 port [tcp/*] succeeded!

If that succeeds and Koha still finds nothing, the network is fine and the problem is one of the five settings above.

A full original record takes a trained cataloger fifteen to thirty minutes. The same record downloaded from a national library and corrected takes two or three. For a library adding a few thousand titles a year, that is the difference between a backlog and a catalog that is current.

It is also the difference between records that follow a standard and records that follow local habit. A downloaded record carries proper authority-controlled headings and a classification number somebody argued about, which is exactly the part that is hardest to produce well in-house.