[Pipet Users] Is Narval the AIL Piper may need ?

Nicolas Chauvat Nicolas.Chauvat at logilab.fr
Mon Feb 12 06:48:08 EST 2001


For once I went off-line for the week-end and when I came back my inbox
was full of nice ideas about Piper layers and uses, and why Narval (or
parts of it maybe) should be integrated to it as an Artificial
Intelligence Layer.

Short answer is: hold on. Long answer is: sure, why not! But I'd like all
of us to be clear about what we're doing. As the history of the Piper
project shows, you all seem to be pretty good at abstracting things out.
So let's sit down a bit, forget about what we have today and what projects
the current code base inherits from.

Two years ago when I started Narval (before I could get the others to hop
in so that we get it to work in less than 5 years), I was also thinking
about a visual shell and discussing it with collegues. Visual-shellish
ideas indeed sliped into Narval's design, but it's *not* a visual shell in
the sense Piper appears to define it and doesn't intend to become one. On
the other hand, it could very well be made to interact very nicely with a
visual shell like Piper and share code and other things.

To get back to my previous point, I'd suggest forgetting about the current
code base and write down a rather precise description of what we want
Piper to do (you know, requirements :-). The ideas of piper nodes that
publish their capabilities, P2P capabilities for communication and
resource discovery, extending the principle of pipes, visual programming,
etc. are the real goal(s) of Piper. It could be that choices were made
when first merging the intial projects because anything else would have
been impracticle (I'm thinking about the mapping one person / one layer
here) and that those choices are actually not so good when we look at what
we want to get to. (Again, I can't say for sure, one way or the other, so
if someone has a clear opinion on that, please speak up).

I'm willing to help here, so I'll read anything about Piper you can point
me too and I'll try to send to the list a write-up about Narval
achievements, hopes and dreams (have quick look at
http://www.logilab.com/press/linux-expo2001 as a starter). You'll have to
wait before I can find time to try the code, and at this point, I'm not
sure it's mandatory. Then I'd suggest we collaboratively write everything
down and post it on the website so that any passer-by can comment. I'd say
that's what we started doing this week-end on the mailing list and with
the Wiki, but maybe staying too close to what exists and not what we're
trying to get to. Once we get that document, I'm sure the questions about
"Should Narval be Piper's AIL?", "I thought about adding a Genetic
Programming module, where do you think it would best fit?" will get rather
obvious answers as we'll figure the implications right away.

If you guys agree with me on this need of discussing and writing down the
goals of the Piper Project before continuing further with the design,
please give me (or Logilab, so that Alexandre can use it too. Name it
after Narval's mascot: ornicar :-) an access to the CVS and I'll check in
the very first version of that document so that we can all start building
on it.

If we were to build a web browser or some common piece of software, the
roadmap would be crystal clear. Here we're dealing with something that
looks like research, or at least pretty innovative stuff. If we don't have
a clear understanding of others' ideas and do not share the same vision,
chances are the road will be winding. Free software or not :-)

Nicolas Chauvat

http://www.logilab.com - "Mais où est donc Ornicar ?" - LOGILAB, Paris (France)

More information about the Pipet-Users mailing list