Comments on: Distributed Version Control and Library Metadata

By: Idle Hands Are The Devil’s Plaything « thesecretmirror.com

Idle Hands Are The Devil’s Plaything « thesecretmirror.com — Tue, 07 Jun 2011 05:50:26 +0000

[...] (ahem, vote for mine, if you're so inclined): Building on Galen Charlton's investigations into distributed version control systems for metadata management, I offer a prototype system for managing archival finding aids in EAD (Encoded Archival [...]

By: Jon Gorman

Jon Gorman — Mon, 21 Jul 2008 16:11:44 +0000

Ah, interesting. I haven’t done as much with the distributed version control systems as the…well, traditional ones I guess you would call them. Obviously I need to play around with them a little more.

By: Jakub Narebski

Jakub Narebski — Sun, 20 Jul 2008 21:27:38 +0000

@Jon Gorman: Git does rename detection based on similarity score, so it should not matter whether one person caled file ‘foo’ and second ‘bar’ when creating almost the same file: they should merge correctly.

By: Jon Gorman

Jon Gorman — Mon, 30 Jun 2008 15:41:14 +0000

Love the idea and I’ve been arguing for something similar for a while. (I’d love a change history for our records).

One other possible issue that popped into my head this weekend is the possibility two different records get created for the same resource at about the same time in two different libraries.

Say Susan at X makes a record for “How to conquer the world in thirty days; by Khan”. Bill at Y got the book around the same week or so and Susan hadn’t actually submitted her file (1234.mrc) upstream yet. Bill doesn’t find any existing records and creates a new one as well (24444.mrc). Now we start getting fragmentation.

The way most dvcs work on is based off filename I believe. This could be an issue if each institution assigns a record for the same source with different conventions. We could try to use an identifier with ways of correcting for conflicts with identical isbns in different books, etc.

This makes me a little worried we’d have to do some custom tweaking of a dvcs, which might be nightmarish. I suppose we’ll probably need to create custom searches and indexes anyhow from the system, but we can reuse some tools for that.