Re: Fw: Divide and Conquer

[Alex Rousskov <rousskov@xxxxxxxxxxxxxxxxxxxxxxx>]

Furthermore, in order to move to a conclusion faster, I think we
should start with what you propose last: a good set of specific
examples that rules language can handle. I will try to contribute
some real-world ACL examples soon.

Thank you,

Alex.

On Wed, 10 Sep 2003, Lee Beaumont wrote:

> I am interested in the IRML vs. P debate going on here. With the
> proposal of P I believe Alex has tackled two problems at once.
> Perhaps if these are separated, more clarity can lead to more
> progress. The two problems I see being confounded and addressed are:
>
> 1) What style syntax should be used in the language?, and
> 2) What expressive power should be allowed.
>
> These issues go to the root of the requirements for a rules
> language, and also involve some parochial preferences.
>
> Regarding syntax, people naturally have a prejudice toward either a
> verbose (e.g. COBOL and XML) style of syntax, or a preference for a
> terse style (e.g. C and its derivatives, like java and P). My
> preference is for the terse "C" style, and I believe the success of
> "C" gives some broader support to that preference. Perhaps this
> debate over syntax style can be put on a more objective footing by
> examining:
>
> 1) Ease of writing,
> 2) Ease of (human) reading,
> 3) Availability of editors, visulaizers, and verifiers.
> 4) ease of machine interpretation, including speed, and size
>
> Related to expressive power, I believe the debate needs to focus
> first on the types of rules that need to be expressed. For example,
> can the WG establish a "working set" of "pseudo code (or structured
> English) rules that need to be accommodated. This might include
> access control , Anonymization, local preferences, natural language
> translation and other examples chosen from the OPES documents. Then
> this working set can be expressed (or not) in IRML and P and the
> results compared.
>
> I hope this is helpful
>
> Lee Beaumont
>
>