Tim Spalding has announced thingISBN on the Thing-ology Blog. It works very much like OCLC's xISBN service that I've mentioned here several times. He also has a very interesting option that allows you to see which ISBNs thingISBN has in common with xISBN, and which ISBNs each service has the other doesn't. Obviously, with a billion holdings vs. LibraryThing's three million, xISBN has more ISBNs in it, but LibraryThing does quite well, especially for paperback ISBNs.
One of the main differences in the services is that thingISBN relies on explicit links across editions/expressions, while xISBN computes everything based on the MARC data. Of course, the MARC data has ways of bringing together (and separating) these, so maybe the difference is more an interface issue than a real one. We also go to great lengths to make sure each of the ISBN clusters is distinct, so that you always get the same cluster no matter which of the ISBNs in it is used. I talked to Tim about thingISBN, but forgot to ask him about that.
--Th
Thanks for mentioning it.
Yes, it should always give the same ISBNs. (However, it's based on live data, without any cacheing, so an ISBN could be added in between requests.)
On the differences, while it's true that "thingISBN relies on explicit links across editions/expressions" and xISBN relies on algorithms, another difference is in who's creating the links/algorithms. LibraryThing's links are made by "regular people," xISBN's by "authoritative people." I should add that an algorithm does make a stab at associating new ISBNs, but it doesn't go much beyond comparing title and author strings with the punctuation removed.
Posted by: Tim Spalding | June 15, 2006 at 16:33