specs

Laurent Martelli martelli@iie.cnam.fr
05 Jan 1999 00:54:10 +0100


>>>>> "Brian" == RE01 Rice Brian T EM2 <BRice@vinson.navy.mil> writes:

    Brian> that would allow us to use Lisp to simulate the arrow
    Brian> system, but the limitations of Lisp might be cumbersome for
    Brian> some of the things i would lilke to express in Arrow.
    Brian> maybe it could be used for a portable description of graphs
    Brian> or something.
    >> What limitations are you thinking of ?
    >> 
    Brian> at a simple level, i'm just talking about the idea of
    Brian> expressing a graph in a linear text-based format: it's just
    Brian> not intuitive.  

I am afraid I can't agree on this :-(. The linear text based format is
just one way of representing the information. And to me it does not
matter. What matters is the semantics behind the syntax. I can imagine
an interface to Lisp which would enable you to handle "graphical"
items such as arrow instead of linear text. 

And in many domains, arrows are probably not the best to visualize
information. 

I advocate the complete separation of semantics and representation of
information. One piece of information can be visualize in many
different ways, depending of the preferences of the user.

    Brian> another thing is that some of the abstraction operators for
    Brian> which i'm developing explanations do a lot of
    Brian> 'cross-cutting' that would result in an overload of
    Brian> identifiers.  

Overloading of identifiers looks like a syntactical problem to me. And
I think I have already expressed my opinion on syntactical problems. :-)

    Brian> what i mean is that the Lisp system which emulates an Arrow
    Brian> system would probably have to be more complicated, due to
    Brian> the inherent difficulties in simulating infinitary
    Brian> structures of arrows, which i am starting to realize wil be
    Brian> quite common, as well as some prolific arrow structures
    Brian> generated for even simple systems.

Infinitary structures means infinite quantity of information. And this
is a problem anyway if you consider that we only have finite amunt of
memory. 

And if you mean that prolific arrow structure might lead to
performance problems, let me tell you something : I think the same
about performance problems than about syntactical problems :-) (it
comes after semantic related problems).

Laurent