[cap-talk] Don't understand capabilities

[Marcus Brinkmann]

At Sun, 29 Oct 2006 18:58:05 -0800,
"Mark S. Miller" <markm at cs.jhu.edu> wrote:
> 
> David Hopwood wrote:
> >> They are no more or less so than "objects" in the object-oriented
> >> programming sense.
> 
> Marcus Brinkmann wrote:
> > Why do you think so?  Certainly, the human mind has a great capacity
> > to identify and abstract objects in the real world.  Furthermore,
> > "operating on an object" seems also to be a concept inherent to the
> > human mind, as the passive voice shows which is part of our language
> > (have to check with a linguist if it is universal, but it's hard to
> > imagine exceptions).
> > 
> > I don't see how a similar argument could be constructed for
> > capabilities.
> 
> David is not talking about "objects in the real world" (OITRW) but about 
> '"objects" in the object-oriented programming sense' (OITOOPS). There is a 
> chain of arguable similarity here:
> 
> OITRW =similar-to=> OITOOPS =virtually-identical-to=> capabilities
> 
> If OITOOPS are understandable because of their similarity to OITRW, and if (as 
> David and I believe) OITOOPS are virtually identical to capabilities, then one 
> should expect capabilities to be just as understandable as OITOOPS because 
> capabilities are just as similar to OITRW.
> 
> What salient differences do you see between capabilities and OITOOPS?

I think the confusion I am causing (and I am sorry for this, because
it is not really necessary, and it is a minor point anyway), is
because I am trying to attack the issue of "understandability" from a
completely different angle.

I am not talking about rational analysis and synthesis.  At that
level, the basics of capability theory are downright trivial to
understand, although many of their applications in high level real
world systems are decidedly not.

What I am trying to talk about is a level of understanding something
that works below the level of rational thought, in the inaccessible
inner workings of our mind, inherent to our body.  Not much is
understood (in the rational sense) of these inner workings at
scientific certainty, but it is a highly active research topic, and of
course we know a lot at the factual level from our own experiences
(and the documented experiences of mankind).

For example, it is well-funded to assume that even animals process
brain structures for elementary "arithmetics", comparison and even
additions of small numbers.  This is not a digital CPU inside us, it
has rather floppy precision (the capacity decreases as the numbers
increase).  Research in this area has put into question the structure
of the mathematics curriculum we teach our children at elementary
school.  This is a peculiar issue: The rational construction of
mathematic theory collides with the natural development of
mathematical abilities at a young age, and we can all see the result
of that (many people who leave school are scared of mathematics for
the rest of their life).

To teach something well, or to make something easy to use, I believe
it is not sufficient to make it accessible to rational thought,
although that certainly helps.  I believe it must also be made as
"compatible" to our organic intellectual structures as possible
(without corrupting the theory of course).  This is not an easy task,
and I am not in a position to make suggestions, but it is worth
thinking about it, I believe.

I think is also instructive in this context to look at the history of
physical theories and their teaching, in particular theory of forces
(field theory) as well as relativity and quantum mechanics.  Feynman:
"I think I can safely say that nobody understands quantum mechanics."
Think about that statement.

I do not want to ramble, so let me finish here with answering your
question: I still believe that from human experience designation and
authority are separate concepts.  Thus, the restriction to naming in
capability systems is artificial.  Unfortunately, it is also a
critical restriction to the successful construction of well-behaving
systems (it is not clear to me if this is inherent, or a limitation of
today's state of the art---I am leaning towards the former, because in
the end everything boils down to information processing, in which all
constructs such as objects, identity, etc are completely artificial).
This is one of the things that can easily confuse a newcomer, even
programmers, depending on their background.

Thanks,
Marcus