The Herd of Kittens - Library Design Notes

This is just a collection of ideas and observations I've had or collected regarding an online library catalog. It will be updated over time; feel free to email me any additional suggestions.

Data Gathering

Most barcode readers are easy, you just get text. However, while you should clearly always scan the EAN, you should also scan whatever code is on the back cover - while you must have the EAN to do useful lookups, having whatever unique backcover code is available lets you do faster "checkout" of a book. (I am surprised that Wordsworth doesn't do this - the cashiers open the book to scan the EAN.)

"Existence" is not the only feature of a book: location is also useful. One simple way to record this is to tag the shelf itself with a barcode, and scan that when placing the book, or when doing the initial bulk load. In general, verbs and nouns can be printed as barcodes on pieces of plastic or cardstock which are kept handy in the library, to avoid gratuitous keyboarding. (Note that this means where the book "should" be, as opposed to a temporary location like "signed out" or "in the bedroom".)

When a book is returned, scan it, and the software should be able to acknowledge the end of the loan, as well as give a hint as to where the book should be located. Text-to-speech code could make this even simpler - it would allow mounting only a scanner, a speaker, and a few barcode cards at the shelf.

Library cards should also be barcoded, of course for convenience.

MARC card catalog records appear to be of varying quality. Some heuristics are useful but hand-editing will always help.

An XML DTD for MARC records is available. This makes it easier to include MARC records as an element of a general book record, or at least to reference them from library-specific records.

It will eventually be interesting to include image scans of the spine and cover of each book, possibly for a VRML or otherwise graphical library shelf.

For reference, MITSFS has little useful data online - only author, title, and existence/condition. A system such as this doesn't appear to fit their mission, either, though there are certainly interested individuals there. Gathering the data would be a phenomenal task, as they have roughly 21,000 titles1 (as of 1999q2, when they had acquired 1,000 new titles2 in the past year.)


  1. "21,000 titles" from, via zephyr, 19990330 

  2. "1,000 new titles" from, via zephyr, 19990330. Also noted that they have 60,000 volumes, and had been collecting for just short of 50 years.