Re: a seriously naughty thought...

Hi, folks; thanks for all the very interesting comments on
my Naughty Thought of an all-Python ARK.  Sorry to be slow
in getting back to this thread...  I think it would be best
if I comment all in one message, so here goes.

Martin Andrews writes:

> Funny you should mention this. I had just embarked on
> writing my macro-language for cfengine as pure python. ...

and later Ronan Keryell pipes up:

> Do you mean PCFengine ? :-)
> 
> http://www.lit.enstb.org/~keryell/projets/PCFengine/expose/2002-09-12/

The stuff in those slides is v interesting; do you (or
Martin) have real code that anyone can look at?

Ronan also says:

> Anybody interested by joining the effort ? Could be
> interesting to discuss at LISA for example.

I for one will be at LISA (including the configuration
workshop on the Sunday) and would greatly enjoy talking with
you about these things!

Martin also said, in a later message:

> ..., I am happy to keep my data and code complete
> divorced. I like the example of specifying the domains via
> XML.  That makes sense. Defining them with python code
> would be messy (and I would not feel safe letting the
> other admins add addresses). But I would be happy if the
> definition of those prototypes was done in python.

Hmmm... that is a possibility I hadn't quite thought about:
having "real" objects be XML and having "prototype" objects
be Python.  (The current distinction between real and
prototype objects was supposed to be a temporary hack, for
simplicity's sake, until the distinction started to hurt --
which it hasn't so far :-)  However, I am a little worried
over how you can have "data and code completely divorced"
and still have an object-oriented system (for any definition
of "object-oriented")...

Or, given how Jonathan reminded us of some of the more
lovely features of ARK as it is -- fairly compelling stuff,
dude... -- I guess we could have *two* inheritance trees for
a given thing: one for its attributes (current ARK
lists/tables/etc.) and another for its methods (somehow all
in Python, or maybe not). ???  Current ARK methods are
pretty terrible -- generally hard to pass them arguments,
and hard to run them for any purpose other than their
side-effects.  It's surprising they work at all :-)

As to the XML thing, though I've (obviously) been messing
with it for a few years, I've tried to steer well clear of
the XML Hype Festival (which has mostly failed to deliver,
so far).  At least in ARK as it is now, XML is just
*syntax*, for something not that complicated.  I'm glad to
hear some enthusiasm for it on the ARK side, though, as it
suggests the original choice wasn't that bad :-)

Jonathan, no, don't give up on your More Document Less
Program theories!  We're still just talking...

(Gotta run.)

Will


-------------------------------------------------------
This SF.NET email is sponsored by: AMD - Your access to the experts
on Hammer Technology! Open Source & Linux Developers, register now
for the AMD Developer Symposium. Code: EX8664
http://www.developwithamd.com/developerlab