Re: [Gnu-arch-users] Re: arch roadmap 1 (and "what's tom up to")

[Tom Lord]

    > From: Miles Bader <address@hidden>

    > John Kinson <address@hidden> writes:
    > > Why not Python?

    > Python doesn't seem to meet the goal of a svelte embedded language (or
    > at least the standard implementation doesn't).

    > I can't really speak for how well it matches Tom's other criteria, as I
    > don't know that much about it.

    >    $ size /usr/lib/libpython2.3.so.1
    >       text           data     bss     dec     hex filename
    >     924082         146624   12276 1082982  108666 
/usr/lib/libpython2.3.so.1

That is not how I would measure the size of the implementation.

I'd be more interested in (but I'm not asking you to report on because
I'm not really all _that_ interested in):

    ~ the size of the smallest usable subset of libpython2.3
      that can be linked against, sacrificing inessential built-in
      functions and types.

    ~ the existence of or clear possibility of an implementation
      of Python which is optimized for simplicity and size

I'm not _really_ interested in those numbers because Python "fails" my
criteria for much simpler reasons.  The most fundamental of those
reasons is the lack of a general orientation that "code == data" which
implies poor to non-existent support for macro systems and other code
transformations.  I'm sure I can find many other little glitches in
Python but why bother?  Stop there.  That's a show-stopper.

That said, Python is an interesting language that came on the scene
when many more programmers were skeptical of high-level languages than
are today.  I think that history will show it (along with Java and a
few others) to have been a good _teacher_ to a generation of hackers.
Those languages led people much closer to understanding and being
comfortable with lisp family languages because they both resemble
substantial subsets of a lisp family language.  Programmers
experienced (and are experiencing) the resultant increase in
programming power; when they get to the pinnacle of the Python or Java
experience, they encounter practical difficulties and the ones who are
bright enough to recognize that need only _ask_ a good lisper for his
approach in lisp to learn of the gap between Java and Python and Lisp.
(In practice, those questions often turn into flame wars but sometimes
there is actual light in addition to heat.  Hearts and minds are
always won one by one.)

I sometimes like to tell people that Python, as the name suggests, is
actually a _joke_ and the punchline comes when, after some future
release, the Python community realizes it has essentially
reimplemented MIT Scheme (a dialect of Scheme extended dynamic
features that resemble (with a few twists) the object system of
Python.

-t