coherent states

[Matt Miller]

On Tue, 23 Jun 1998, David Manifold wrote:

> On Mon, 22 Jun 1998, William Tanksley wrote:
> 
> > ...I see two dangers:
> > first, you might spend all your time working with our little Tunes community
> > developing our little Tunes vocabulary and then only when we finally release
> > Tunes do we discover that our little vocabulary doesn't match reality (the
> > solution to that might be getting flames from the Usenet); the second
> > danger is that you might spend all your time making the vocabulary
> > perfect and never attain the working system.
> 
> In my opinion, perfecting the vocabulary is the same as making the working
> system.  At least, when the vocabulary is complete, the bulk of the work
> that we need to do for Tunes will be done.  Following a specification is
> easy, once you have one.  Getting the specification in a preliminary
> state, which we are doing right now, seems harder.
> 

I suppose I agree with this.  However, in an earlier message you indicated
that "easy to use" would be part of the specification.  I assume that when
you talk about specification here, you are, in fact, talking about
something much more, well, specific.

Since I'm writing anyway, the only way that I see to make a system truly
"easy to use" is to have the system anticipate the desires of the user.  I
don't know how you go about implementing that.   The biggest hurdle is the
first use problem.  How does the computer interact with the user in their
first exchange?  I once read an article by Jerry Pournelle in BYTE which
indicated that he thought the mouse was counter intuitive since it moves
in the XY plane and the screen pointer moves in the YZ plane.  How can you
anticipate all the varied demands of every user?

Matt