BAM-Portal

» see bam-portal.de as it lived, on the wayback machine

~ the eulogy ~

The acronym is German and it names the sectors it joined: Bibliotheken, Archive, Museen. A library catalogs a book by author and title. An archive describes a box by provenance and by the office that produced it, while a museum record starts from what the object physically is and what it is made of. None of those are describing the same kind of thing, and getting them to answer one search field is the whole story of this site.

It went online in 2006, after a run-up that started with a data-warehouse prototype in 2005 and money from the Deutsche Forschungsgemeinschaft that ran until the middle of 2007. Five bodies carried it after that: the Bibliotheksservice-Zentrum Baden-Württemberg and the Landesarchiv Baden-Württemberg did most of the operational work, with the Stiftung Preußischer Kulturbesitz, Digicult Verbund and the Technoseum in Mannheim alongside. The index was Lucene. The sources were the southwest German library union catalog, the state archives of Baden-Württemberg and Hesse, the Bundesarchiv, municipal archives and several million museum object records.

By the time it closed it held close to 70 million records from more than 50 institutions, which made it the first cultural portal in the German-speaking web to hold that much across that many kinds of collection. It shut on 30 June 2015. The partners had spent the previous year agreeing that the Deutsche Digitale Bibliothek, which was permanently funded and nationally scoped, now did what BAM had been built to prove was possible.

~ the dossier ~

~ merging inventories that disagree ~

Federated search over heterogeneous metadata is a problem that looks solved from outside and is never solved from inside. Every source hands you a different schema, a different authority file for names and a different idea of what a date is, and a union index has to flatten all of it into fields a person can actually type into. Flatten too hard and an archivist tells you the results are meaningless. Flatten too little and a visitor gets 40,000 hits with no way to narrow them.

BAM's answer was a warehouse rather than live queries against every partner. Harvest the records, normalize them centrally, index the result with Lucene, accept that the copy is a day or a week behind. That was the correct engineering call in 2005 and it is still the correct one, which is why it is what every successor built too.

~ closed on a schedule ~

Infrastructure like this usually dies when a grant cycle ends, and it leaves its records stranded on a server nobody is paid to keep. BAM got the other ending. It proved a thing could work, was replaced by a permanently funded version of itself, and switched off on an announced date with its data already moved.

The consortium went further and formed a BAM group inside the Deutsche Digitale Bibliothek, announced in June 2015, so the partner institutions had somewhere to go on the day the portal stopped. The records outlived the address. For a search engine that is the good outcome.

~ the part nobody outside Germany saw ~

This one had no audience beyond people who needed it. A German-language union index over state holdings is not something you discover, it is something a librarian tells you about. From Ankara there was no reason it would ever have crossed my screen, and the thing I find interesting about it now is not the collections but the plumbing, which is the same plumbing anyone who has tried to merge two inventories with different primary keys ends up writing by hand.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »