[self-interest] On "Meta," with a trick question at the end.

[David Ungar]

Well....if Randy is going to jump here, I could say that I agree with 
him. It is all about
your purpose. For Self, we picked a set of things that were useful in 
regular programs and
conceivably implementable, and encapsulation-respectiing, based on 
our idea of what
object-oriented programming could be.

Then there was a bunch of other stuff that needed to be there for the 
programming environment,
that took an object's destiny away from it. That seemed like the meta 
stuff and went into the mirrors.

- Dave



At 4:23 PM -0800 1/16/01, Randy Smith wrote:
>Don't forget to look at the trick question at the end
>of this note. Meanwhile....
>
>
>>  > > what is meta-programming?  >
>
>Jecel gave 3 good examples, including a Self program.
>
>>  > This program is *about* a program -
>
>..and this to me is the key word:  "about."  A meta discussion
>is about a discussion.  A meta program is about a program.  A
>meta X is about an X.
>
>Now one must use some interpretive framework to decide what
>something is about or even if it is about anything!  So the
>question of the degree to which something is meta is one of
>interpretation.  In other words, rational people may disagree
>about the metaness of some object.
>
>For example, information about the representation of an object
>is normally considered meta.  The number of slots, the names of
>the slots and so forth.  Now consider:  is the size of a
>collection a meta notion?  It does not depend on the
>representation of the collection, and is at times something you
>want to know when working with the thing.  Hence some might
>consider a collection's size a rather everyday, non-meta
>concept.  Others might say it is meta, as it is clearly about
>the collection.  Obviously in Self we take the former view, and
>let you directly ask the object for its size (in Self we use
>mirrors for the meta stuff, like slot manipulations).
>
>So the choice of trying to put "meta" aspects into a special
>world of mirrors is a choice to confront a mess of vague and
>arguable decisions.  However, we figured it was a practical
>factoring, even if it was an uncharacteristically non-Self-like
>to introduce unneeded distinctions.
>
>Hey, how about the metaness of the following, which refers to
>itself and is therefore meta?  "If this sentence is true, it is
>not meta, otherwise it is meta."
>
>Is it an empty trick or does it reveal a problem with the
>meta/nonmeta distinction?
>
>--Randy

-- 


     David Ungar
     Sun Microsystems Laboratories
     (650) 336-2618