A strong content language for FIPA
----------------------------------

Summary:
This proposal is for a workplan to specify a strong content language
that reflects the best of FIPA experience and which may significantly
improve inter-agent interopability.

This is a first announcement, and solicitation of interest in such a
work programme. If your interests are congruent with the following
description please let me know, and please consider contributing both to
the workplan and to the work itself.

Background:
The current story within FIPA relating to content languages is a little
weak on a number of counts:

Not having an `official' content language seriously weakens the case for
developers considering FIPA compared to other approaches (e.g. KQML/KIF,
OAA).

A content language is required by particular work programmes within
FIPA, both today and in the future. E.g., agent management requires a
language in which to express agent management actions and concepts.
Currently this is done in SL0, but ...

SL is close to being a de facto content language for FIPA. However, SL
has serious problems when considered as a content language: not the
least of which are an uncertain semantics and a lack of engines capable
of processing interesting SL expressions. Both of these are not easily
fixed.

It is important to be able to easily represent and manipulate many kinds
of expressions, including `non-logical' entities such as Java objects. A
string-oriented content language will always be weak in capabilities for
handling such real-world entities.

In addition, there is a percieved need to be able to express
higher-level agent interactions such as shared agent plans. 

Requirements:
We see the following as essential prerequesites to any content language
activity:

1. A 1st order logical semantics, preferably Horn Clause Logic (i.e.,
Prolog oriented semantics)
2. Able to represent actions, including ACL speech acts (the strong part
of the strong content language)
3. Able to represent shared plans, agreements between agents and other
entities and able to represent policy statements and/or constraints.
4. Compatability (where possible) with the existing semantics of
existing ACL
   performatives.
5. Core semantics and structures expressed in an abstract way to permit
maximum flexibility of representation within particular architectures
5. Able to represent conversations and concepts shared across several
agents
6. Able to represent and manipulate multiple ontologies and symbols of
different ontologies
8. Offer a framework to permit additional third party content languages
to be viewed as restrictions and special instances of the strong content
language.

Proposal:
To initiate the development of a specification of such a content
language. Given the proposed Prolog-based semantics it should be
possible to constrain a transition of a specification to Experimental
status based on real implementations.

This would be in the context of a specific TC set up to focus entirely
on this activity. A reasonable estimate of the required effort is 8
meetings of the TC, together with some off-line editing of the proposed
specifications.

Relationship to existing FIPA specifications:
A strong content language would fit as a member of the FIPA content
language library. At this point in time it would not be appropriate to
mandate the use of such a hypothetical CL by other FIPA work programmes;
however, we would hope that the resulting specifications would be
sufficiently attractive to other FIPA members to use within FIPA and
outside.

Partners who have expressed willingness to pursue this work:
Fujitsu