[open-bibliography] Announce: Open Bibliography (JISCOBIB)

Fri Jun 25 12:02:45 UTC 2010

On 24 June 2010 17:57, Tim Spalding <tim at librarything.com> wrote:
> A few comments.

[...]

> I am, however, interested in *just what data is produced, or opened*
> in both extent and detail. If this project produces a linked-data
> representation of the BL's library records, with all MARC data
> preserved in some way—great. There's a lot that can be done with that.
> If it adds data to that, from another source or through internal
> analysis, double great. If it produces some lossy representation,
> either in extent or detail, which can't be used for cataloging or to
> add information to traditional catalogs, that becomes a lot less
> interesting to me. I'm simply unclear which is happening here.

The aim would be to preserve all info and I'm pretty sure efforts
would be to made to ensure availability of the the source MARC records
too (so if people don't like the RDF-izing you can do it differently).
However, exactly what we do does depend on the source libraries.

>> We have two very sets of use cases. Rufus and Ben will be working with key
>> libreary catalogues (BL and Cambridge). Here we can expect some records to
>> be very complex and the expectation of usage very complex and varied
>
> So, is BL data going to be opened up?

For clarity: the BL (and CUL) are *not* committing to open up *all*
their data at this stage -- it is likely that CUL will be able to
release more than the BL by the way (both CUL and BL are "copyright"
libraries in the UK though BL's holdings are larger [1]).

Discussions are going on right now to determine the extent of the
release -- I should note that this is, to an extent, a prototype
project and about showing institutions what the benefits of open data
might be. If it is successful I hope we will see more extensive (and
perhaps complete) releases of data.

[1]: http://www.rufuspollock.org/tags/eupd/

>> That sounds useful. I don't quite get the point of keeping your open data
>> closed.
>
> I mean merely that the data is already open through others. Open
> Library and ‡Biblios provide access to the same data. So we don't have

Aside: How do you actually get data out of Biblios (especially in
bulk)? I've never been able to work out how, which is why the
resources field of the CKAN package remains blank:
<http://ckan.net/package/biblios>. Also where does there data come
from? One often has to be a bit wary of where data got sourced from:
e.g. if you just scrape Amazon you are probably grabbing data from
Nielsen (one of many parties who probably supplying amazon) and
Nielsen run a significant business selling that data and are likely to
have rights in it ...

> a public API to the data because we don't see much of a need. The
> software that gathers and processes it is, however, pretty good.
>
>> "Data licensing" will depend on what the data are. The OKF has tools for all
>> sorts of "data".
>
> This is a two sided question. I'm interested in what OKF
> discovers/decides about the legal state of book-and-article records.

My feeling on book records is caveat emptor -- data rights vary and in
most jurisdictions you're going to need to worry about the rights in
data (see [2] for more). Remember also that issues don't arise at the
beginning, they arise when you really start to threaten some
incumbents business model ...

[2]: <http://www.opendefinition.org/guide/data/>

To deal with this risk, my view is that open data projects should take
an an open street map type approach where you:

a) try to determine and record provenance of data fairly strictly
b) take a "clean-room" type approach on what data you accept and aim
only to take data that is explicitly open

> And I'm interested in whether LT should release its data to OKF, and
> if we did, what licenses could be applied to it. For example, we
> offered to release our Common Knowledge data to Open Library, but they
> refused the CC-Attribution license we proposed.

We'd be happy to accept any open data license, i.e. any license here:

<http://www.opendefinition.org/licenses/#Data>

And it you are looking for attribution Open Data Commons (a
OKF-affiliated project) has just today released an attribution license
for data(bases):

<http://www.opendatacommons.org/2010/06/24/open-data-commons-attribution-license-released/>

Rufus