More on clustering in springalgo

---------- Forwarded message ----------
Date: Wed, 3 Apr 2002 20:31:30 -0500 (EST)
From: cdent@burningchrome.com
To: katy
Subject: spring and such


I talked with Jason a bit and it was clear from our conversation
that the thing to do at this point is just proceed. On Sunday he and
I are going to meet and we'll see what we can straighten out. Our
goal for that day is to integrate out code bases. I'm going to
keep working until Friday and then let him know what I've got and
stop at that point so he can have something static to go from.

Then after Sunday build from there for Wednesday's presentation.

I'm somewhat concerned that his model for doing clustering limits
the "toolness" of the thing we are creating. I've been trying to
create something that is as general as possible given the
following criteria:

- display nodes and springs
- nodes have repulsion
- springs have attraction
- the display has gravity (this turns out to make a -huge-
  difference in getting things to look nice, but there are
  multiple ways to do it, my current method is not very nice)
- nodes may have hierarchy
  - that is, nodes may have children and a node can be told to
    display its children
  - or an entire slice of the hierarchy may be displayed
  - these two can be combined

I believe that Jason wants to put the clustering algorithm
somewhere into the model. This makes the assumption that the
incoming data represents a matrix of some kind, which doesn't
have to be the case: Spring is a very general layout method,
usable for all sorts of stuff.

It seems that clustering should be a pre-processing aspect.