[okfn-discuss] The Four Principles of (Open) Knowledge Development (v0.1)

[Rufus Pollock]

Saul Albert wrote:
[snip]
>>>I think these are very good and clear, the only worry I would have is
>>>that many people have no idea what a package manager is, and perhaps
>>>there are examples of componentized knowledge that are more widely
>>>understood. However, I do see the very real value of the apt-get example
>>>because it is so amazingly well done (apt-get install anarchism!).
>>>Perhaps another example to go alongside it - or another section called
>>>'examples'. This section could look something like this:
>>
>>Good idea again. I wonder if we can think of any more examples other 
>>than the 2 we now have (apt and bibliographies). This also prompts me to 
>>bring up the idea of doing a simple 'hello-world' type knowledge package 
>>-- more of which in a seperate post.
> 
> 
> Well, I had a think about this and an interesting talk with Fabian
> Tompsett last night. I mentioned to him the Open History threads and the
> Open Shakespeare idea, both of which sound good and have been discussed
> interestingly on this list, although from the sounds of it, bootstrapping
> Open History from Closed History will be a challenge :)
> 
> I have one more idea - which is still a bit of a mess of interests, to
> add to the pile:
> 
> We were talking about the Distributed Library Project
> (http://dlp.theps.net) - a project we worked on together with other
> members of the University of Openess a few years ago, it's basically a
> library catalogue that allows people to store books and other catalogue
> entries in their own home, but share a simple library card / borrowing /
> commenting / listing system.

I've just taken a look at its looks very nice. One thing that is common
to this and, e.g., the public domain works database is the the core
metadata db for books/films/music. Each project is then adding on top of
this a bunch of functionality specific to its domain (e.g. borrow/lend
for dlp and commenting/ispublicdomain for public domain registry).

As previously discussed_[1] there are some good metadata dbs for recent
music (freedb etc) but nothing much for film or books.

[1] http://lists.okfn.org/pipermail/okfn-discuss/2006-April/000041.html

So -- and I think this is what you were getting at :) -- we have right
in front of us in the form of metadata dbs (library catalogues minus
lending info) perfect examples of 'knowledge packages'. Moreover they
are knowledge packages with clear 'reuse cases' in the form of the dlp,
pd registry, film recommender etc etc.

> It's pretty limited at the moment - and the code base (and data model) is
> somewhat inflexible. It does, however, work, which is the important bit. 

Indeed ;) -- though in line with good eXtreme Programming practice the
Open Shakespeare package in the repository at:
   http://project.knowledgeforge.net/shakespeare/svn/trunk
is fully working (though limited in functionality)

> We were talking about our aspirations for the project, which included,
> most centrally, a system of emergent categorisation and book-threading -
> based on borrowing patterns and lists (somewhat like del.icio.us and
> amazon's recommendations systems). This kind of system is proven and now
> exists. The other kind of categorisation system we discussed, more
> pertinent to this discussion was bibliographical chains. 

Recommendation is trivial once you have use data (just a bit of bayesian
stats). This is another good example of the reuse/extension 
possibilities (both in data and software terms)

> The academic reference system is great - as discussed earlier in this
> thread, and is a *kind of* example of packaging knowledge. The problems

(just to clarify) by this I took you to mean citation dbs (ala Thomson 
ISI citation index)

> with it are that technically marked-up dependency chains of books -
> mentions and cross-references in bibliographies, are
> 
> a). often an academic publication in themselves, with all the attendant
> closeting of info and copyright issues.

+1. i wanted to get data from thomson to do research and it was not 
going to be easy. However there are open -- though incomplete -- 
citation dbs such as citeseer that do provide their data openly:

   http://citeseer.ist.psu.edu/oai.html

(though as usual note that there is no explicit licence). Note the 
second tar.gz includes citation data along with bibliographic records 
themselves

> b). not technically standardised as far as I can tell

Dublic Core seems to be a rough lowest common denominator as far as i 
can tell for /metadata/ (there is also MARC etc) but there's nothing for 
specifying the references.

> c). not built into easy to use software tools.
> 
> We envisaged a system - probably using existing RDF biblio name spaces
> (I know some exist somewhere...) that would allow dependency-chain or

dublin core ...

> tree-like browsing of book collections through cross referencing their
> bibliographies.

that sounds nice (and there would be plenty of fun things to do such as 
search for cycles or different lengths)

> Adding this kind of navigational system to the DLP would be even more
> interesting as if the books exist within DLP collections, they also
> become a set of geographical indicators that have owners - presumably
> people who have read and can comment on the books.
> 
> It wouldn't be a particularly easy project - I mean, some code is there,
> some use exists, we are probably going to start developing a simple RFID

It already looks pretty reasonable

> library card system in the near future, which would make borrowing much
> easier... along with that we might look at this bibliographical
> navigation system, but like all these projects so far proposed, it's
> a lot of data entry!

One always wants to find some way to automate the data entry process -- 
usually by 'borrowing' data from elsewhere :)

Regards,

Rufus