Response by Gregor Kiczales

Unfortunately, my response isn't yet complete. It should be in two parts, the first part will make some general points and the second part will respond to specific parts of the foil. But the more I read the more interesting thoughts come to my mind.

General Comments

Here is just a quick summary of my general comments:

In what it does say, the foil is good and right (what do you expect me to say). But it doesn't go far enough. It should make more of the case for other kinds of reasons why clients need to change things out from under them. Maybe it needs to put forth terms like ``semantic dilemma?''
It also doesn't go deep enough. Fundamentally what is at stake here is thinking about software as being much more malleable stuff. I believe that in order to do that we are going to need for software to look a lot different. Current procedural languages are (among other things) too low-level for what we'll need. They make far too many accidental committments to render the computation they express sufficiently malleable.
As I argue at the end of my OI talk, I think that fundamentally we are heading for a very different conception of the engineering of computational processes.

In the meantime though, I think the foil and the responses are helping to frame the issues in a way that lets us talk about them clearly. That leads me to be optimistic that this discussion is helping us get to where we want to go in the long run.

Specific Comments

The following are comments on specific points in the foil text. They add to the more general comments above.

There's More than One Kind of Abstraction

Implicit in this paragraph is a basic intuition that I believe is core to this work. Namely, that there's more than one way to use the notion of abstraction. In more detail, I believe that:

That abstraction is an extremely basic and general notion;
so much so that there are many ways to do it.
We in software have, to date, been taking just one approach to abstraction;
an approach that recognizes only its use to deal with more or less detail as opposed to its use to deal with this or that issue. (This is reflected in our use of terms like ``layers of abstraction'' and ``abstract away details.''

There's More than One Kind of Dilemma

This paragraph points up one of the major weaknesses of the current foil. It does an excellent job of motivating the need to expose control over mapping decisions, but it doesn't motivate the need to expose control over actual functionality.

Back to Alphabetical List of Responses

Back to Main OI Workshop Page

Back to OI Home Page

Gregor Kiczales,gregor@parc.xerox.com

(Last Revised March 1996)