VAUCANSON 2 : A Finite-State Machine Computation Platform – VAUCANSON 2

This proposal on one hand consists in the
realisation of a software platform devoted to the computation of, and
with, finite automata, and, on the other hand, is the
follow-up (the number 2 after the name VAUCANSON) of a
project of the same kind, that has been conducted by the same
partners during the last eight years, but without any specific
financial or institutional support -- excepted the cooperation
between the two Paris and Taipei poles initiated
via a PHC project.

It is with the experience of this first phase, with its teachings
and successes, with its shortcomings, its problems and the failures to
which they lead, that we have redefined the scientific goals of this
research project, the technical ways to reach them, the financial and human
means necessary to their achievements.



One very characteristic feature of finite automata is that, in
contrast with many other computing models, most of the properties are
decidable and that the corresponding algorithms can be reasonnably
implemented.
Among other systems for computing with automata, the Vaucanson project singles
itself out by the following objectives.

Genericity.
The platform is meant to reflect and accomodate the richness of the
model and to allow
the implementation of, and the computation with, weighted acceptors and
transducers, without restriction a priori on the label or weight types.

Efficiency.
The platform is required to give high, or at least good, performance;
the genericity should not too much handicap
it with respect to other more specialised systems.

Theory faithfulness.
The programming of algorithms and constructions on automata, should
be as close as possible to their descriptions in a mathematical
language.

Versatility.
The platform aims at fitting the needs of the many different users, and
usage, of finite automata.
From first year computer science students to group theorists, from
academics to software industries for system verification or natural
language processing.
This implies that its features should be accessible at different
levels of generality, that the functions in the library will be
called via an API, via system commands, and, in
order to reach the less programming oriented end-users via a
more user-friendly graphic interface.
Finally, but this is more a mean than an objective, the platform will
be an open source software.

The project of the Vaucanson platform has been launched by the Parisian partners.
Few years later, we understood the importance and necessity of having a Graphic
User Interface, and we have then been lucky enough to be able to
bring in the Taipei partner via an ORCHID project (2007-2008).

With Vaucanson 1, we have now a good prototype of what
we wanted when we started the project.
It supports different types for labels and
weights, it offers two distinct possible implementations for automata,
and most of the basic algorithms for weighted acceptors and transducers are
provided.
Vaucanson 1 comes with a set of shell commands, called TAF-Kit,
that allows to compute automata of fixed type (a dozen of different types of automata are available).
Moreover, a prototype of a
graphic interface, called Vaucanson Graphic Interface, gives access to the same service
as the TAF-Kit.

In the new phase of our project, which we call Vaucanson 2 and
which is the object of this proposal,
we want to keep the same objectives as before, which we know even
better now how ambitious they are.
The efforts will mainly address the
following aspects of the platform.
+ A new architecture of the core of Vaucanson in order to overcome
the complexity in the writing of the library and the loss in
efficiency.
+ A stronger specification of the interface of automata.
+ A systematic action for improving the accessibility of the
platform, its documentation and the communication with other
systems.