Date: Thu, 4 May 2000 14:10:56 -0700
|Subject:||Towards an atomic data structure|
[Responding to Eric Armstrong's letter on May 3, 2000...]
Good questions, those.
I don't favor anything. I am *biased* towards experimentation with WBI simply because Doug is. I don't recall saying that WBI is the solution. It is a candidate, a design point along a path, whatever.
I am biased toward knowledge-centric thinking for two reasons: (1) we intend to build a DKR (read: Dynamic Knowledge Repository), and (2) that's the way I think.
I also think Eugene gave a brilliant layout of his needs for an OHS, that thing connected to a DKR that users bang on to get stuff into and out of the DKR. There is an OSS Java PIM that uses XML as its storage medium. I would imagine that simply adding XLL to it would give it the linking ideas he spoke of, and you would be off with a prototype to play with and expand upon, heading towards that Version 1 we all dream about.
In the end, things we don't talk about here, some of which Rod Welch already has running at his web site, must be pondered. One way or the other, they will be pondered.
Firstly, we start with raw information. That's emails, journal papers, books, movies, diagrams, recorded spoken words, and so forth. Next, we must reduce that raw stuff down to filtered (read: interpreted in context) stuff. There is a whole literature out there on Information Retrieval, part of Knowledge Management. Here, we are talking about ways to build indexes into all that raw stuff. Indexes that make sense for all potential users, and that's the really hard part. If the primary categories in your world view are women, fire, and dangerous things, chances are that I'm not gonna come up with an indexing scheme to satisfy your needs. My world view spans, Newtonian and quantum mechanics, molecular biology, and lots more. Not sure how to map all that to women, or fire, or dangerous things. What is the point of a DKR if nobody can get into it?
Following reduction, there is mapping to a representation system for machine inference. Do the reduction poorly and you don't do mapping at all.
All this stuff about email vs web logs is pretty much surface level (implementation level) discussion that should never go on until you have all the requirements worked out. Eugene gave a series of scenarios that could easily lead to requirements for a prototype. What it will look like when implemented should not be on the discussion table (with such passion) at this time.
In the end, Doug, as I recall, set out to get started on a machine that would enhance human productivity in knowledge related work. He took pains to outline global problems, like population, energy, environment, and so forth, and made it, at least IMHO, very clear that such a tool was of enormous value to our future. I happen to come from a background of believing the same thing. To design anything less, no matter what the first out the door version would look like, would be to completely ignore Doug's message. My experience with my program The Scholar's Companion tells me that I don't have to participate in the design of a first hack that prevents its own evolution.