This has kicked around in several versions over the last 12-14 months. I post it here to get it out of my head and move forward some. Links will be added in turn.
For a while now, inspired by a couple sources, I've been kicking around in my head an idea about knowledge sharing, but in a bit of a diffeent form.
A while back, Phil Agre posted on his Red Rock Eater mailing list a message about a desire he had to programmatically know what othwer user's bibliographies were. Phil is a relentless poster of good links, and once in a while - monthly it seems - he posts a list of those books he's seen cited that may be of interest to one of his topics of interest.
This is a nice idea, in and of itself. Part of what most of the people I know who cut a broad swath in their reading is tell one another what's being read.
At about the same time, I'd begun looking at Rael Dornfest's peerkat code (a localised version of the work he's done for meerkat at O'Reilly). What if these ideas could be melded somehow to become a single tool? What if there were a decentralised way to look at the recorded trails of people 'like me'?
Thus was Bibkat gestated. And there it has lain for the last several months. But here's how I see it in germ form.
I keep a catalog of resources - books, magazine citations, bookmarks, purple-numbered emails, etc. Over time this grouping accumulates and gets more tagged with reference and meta data.
In a controlled (secure) manner, I can then expose this information to the world, or some managed subset of it (known fellow researchers, etc). And in turn I may be able to see parts or chunks of other entities collections.
Each entry would be marked up in some common format - in my mind was something like an xml store or Z39.50 store, both of which require additioal research.
In some ways, it should be oclcKat, except that the information here takes on a decentralised form. As I recall it, OCLC maintais a monolithic data center in Ohio where all records get checked in. In the case of bibkat, however, the information is live, including status (is that bookmark still where I said it was, is that book on Chris' shelf, or was it leant out. If its leant out, is it now in Mike's collection, where I might still be able to get some info out of it?)
Since I started musing about this, I've been idly looking at various similar tools, somewhat languidly. Human-Links (mentioned earlier) comes across as having some of the techniques in place, but to date I haven't gotten it to work, and the security model doesn't seem as rich as I'd like this to be. As well, Human-Links is very document centric, rather than resource centric. Pushing down into documents is part of this, but not the first stage.
I've looked at bibliography software. Some of the software for small libraries comes close to the 'publish your data' idea, but they still lock it away and don't really manage to share it as transparently as I envision bibKat doing.
In the online realm, BookApp seemed to have some of the ideas, but I was projecting my own on it. BookApp does do some nice stuff, including an autmatic tie from ISBN lookups, and some sharing of items on a list basis.
A lot of the P2P networks have ideas that flow back into this, but again, they are largely about sharig documents or media, not sharig iformation about resources.
I've a bunch of recommended reading for myself that flows i with this. I need to get a better uderstanding of OCLC, of Z39.50, of Dublin Core, and of databases that handle any of these cleanly.
I need to geta better understanding of how to implement a security model (and how to flesh out my thinking on security. Thankfully I have some application security thinkers in my midst.
There's a number of bootstrap-like thoughts that could tie in here, including purple numbers for local document reference.
So a few months back Phil posted an idle thought. What if he could easily see what was on other people's shelves and in their bibliographic files.....
Posted by esinclai at June 07, 2002 05:45 AM |