CLIM mail archive
[Prev][Next][Index][Thread]

SWM should be developing not supporting

Glad to see Scott's note.  CLIM is great, but we all 
have things we want to see it do in the future --
current things done better or new things done.
So people like SWM should have quiet time to sit and
develop.  The companies may be a little thin on
support but they'll stay thin unless the customers
give them things to do.

When you do send queries to the vendors you can always
copy them to the clim mailing so people will know what
problems are arising.  And *if you get a solution* please
tell us about that too.  We're all in this together.

In our lab, we have developed and/or are refining the
following CLIM applications:

  The Scientist's Assistant to display full biological
   articles including all text with Greek symbols,
   superscripts and subscripts (S. Gauch and M. Tselman).
   This is a hypertext system built on SGML-encoded text.
   Ultimately it's intended to be intelligent but so
   far it's pretty straightforward.

   Problems encountered:  Subscripts and superscripts not
   supported in text mode, must be drawn.  Handling left
   margin and wrap-around had to be done by hand.  Loading
   in a few thousand words of text can take as long as minute
   or two -- not good.  Once loaded, interaction is fast enough.
   Primarily being developed on a MacFX with MCL2.0B3, with
   some cross-development on SLMs.

 The Diagram Understanding System Inspector is designed as
   a development tool for our diagram understanding system
   so we can follow (see) the connections between our analyses
   (parse trees, objects) and the visible diagrams themselves.
   No serious problems here, but it was developed in ILA 0.9
   CLIM and then had to be moved to SLM CLIM 1.0.  Took a little
   work.  (M. Tselman).  One problem:  Can't yet display even
   90 degree rotated text, something appearing in essentially 
   every data graph in the literature.

 The Lexicographer's Workbench.  This will display text much like
   the Scientist's Assistant, but with the goal of identifying
   new lexical items found in the text to be inserted in the 
   lexicon.  Thus a single item might have Greek letters mixed in
   as well as sub-superscripts, e.g., one pronounced 
   "tritiated alpha D glucopyranose" which has a "3" presuperscripting
   an "H", designating Tritium.  (K. Klimukhina)


 The DNA sequencing package.  This operates on and displays 
   multiple windows of raw chromatography data from which we
   develop DNA sequences.  One problem is that it can take quite
   a while to display a lot of points.  (A. Miller)

Some reflections on our experience so far.  We're not trying to do
too much over X; we've got enough to deal with just getting it 
all working locally.

We really need to allow the user (and us) to easily produce and
dispose of various panes and frames -- in our terminology, multiple
articles in view on the screen and multiple views of the same
article.

Performance:  Needs a lot of work to be really viable.  We were
impressed by the fact that the DNA program worked a lot faster on 
our Mac FX than under Allegro on an unloaded Sun 4/390!
I think this says a number of good things about MCL and ILA's CLIM.
What it says about Franz is not settled yet.  We'll see how they deal
with the code we sent them, maybe they can find some things to
speed up.

Since we're an AI lab and are dedicated to putting some intelligence
behind our applications, it make sense to work in CL/CLOS/CLIM.
But delivering an application means a responsive interface.
We're not commercial so we may be forgiven if our stuff is not
blindingly fast, but if *too* slow, even we're in trouble.

The long term hope is that CLIM will be deeply integrated into CL
so that it will perform.  I for one would be happy to work with
any consortium(a) that form to make the CL/CLOS/CLIM community
a stronger one.  NSF has invested a lot of resources in our lab,
so I don't want to disappoint them.  We need to build solid systems
on solid, standardized foundations. Lisp is great (and Norvig's book 
isn't going to hurt either!).

Sorry, if you weren't expecting such a long note, but Scott's
note inspired me.  Wanted to write something above the code level.

bob futrelle


   Prof. Robert P. Futrelle
   Biological Knowledge Laboratory
   College of Computer Science  161CN
   Northeastern University
   360 Huntington Ave.
   Boston, MA 02115

   (617)-437-2076
   Internet: futrelle@corwin.ccs.northeastern.edu



0,,
Main Index | Thread Index