Re: Emacs vista build failures

[Barry Fishman <barry_fishman@acm.org>]

tomas@tuxteam.de writes:
> On Mon, Jul 28, 2008 at 09:00:19PM -0400, Barry Fishman wrote:
>>                           Otherwise I am afraid that the "make GNU
>> look like Windows" people will bury us in the sort of fragile C++
>> monoliths like Firefox, that leave most everyone out.
>
> Note that the basic architecture of Firefox is very much parallel to
> that of Emacs: a C[++] core and a scripting layer on top to write most
> of the user functionality in (Javascript + XML). Not that I am a big fan
> of Javascript (even less of XML), I'd take Lisp over it any day, and the
> sheer bloatyness of Firefox gives me the feeping creeps, but just to
> point that out.

I may be just repeating wrong information, but doesn't Windows (at least
historically) use Basic and assorted data files as its extension
facility.  I assume Vista is moving toward .NET.

As far as Firefox is concerned, I see things like:
   http://developer.mozilla.org/en/docs/Building_an_Extension
It seems to be more a patchwork of extendable areas, rather that
what is going on in Emacs.

I think of Emacs as a lisp program which uses C components to improve
performance and connect to OS facilities.

I think Python has done the best job of doing Emacs style things.  It is
even developing Pyrex, a typed python subset to C translator, to avoid
some of the fragile C macro hacks used in Emacs, Guile, and other C
based Lisps.  It is even REPL based.  But it doesn't have the extendability
and flexibility of Lisp.  Unlike programing in Lisp, Python has the feel
that you are talking down to a small child rather than a peer (child or
adult).

>> From what I understand, Emacs will be moving to Guile (or at least a
>> Guile updated to meet its needs).

Miles Bader <miles@gnu.org> writes:
> This was sort of a vague goal a decade or more ago, but seems pretty
> unlikely to actually happen.

Scheme is going though a painful growth spurt (via R6RS), and its
probably not stable enough to consider at the moment.  Common Lisp has
been excluded by RMS, although some people are going ahead via CL-EMACS.
I assume he feels that its complexity would reduce the hacking community
working on it.

I think there is a continuum from "limited but simple" to "powerful but
complex".  Commercial applications can flourish working at the
extremes, limited but simple for the user, powerful but complex for the
internal developers.  I think free projects really need to provide access
to the middle of the spectrum where hacking can be a incremental
process.

The ideal would probably be to support the large numbers of people
satisfied with the simple but limited area, but with support for more
productive levels of "powerful but more complex" to keep a large
development community happy.  To me this requires the ability to
dynamically build clean application specific sub-languages in the way
only Lisp can do.

My point is maybe some effort needs to be put on how Emacs ideas and
shareable code can be used in a more general framework for building GNU
applications.  Guile, GDK/GTK seem to be steps in the right direction,
but only if they could at least theoretically be used to build an Emacs
that runs at least as fast as it runs now.  But the Gnome juggernaut
seems to be pulling in a different direction.  One where Microsoft is
leading the way.  I just don't want it to overpower Emacs!  I don't want
my frame layouts to be saved in dozens of
"~/.gconf/apps/emacs/**/%gconfg.xml" files!

My own ~/.emacs file based setup has a growing "forward into the past"
section where I roll back new default features that I find better living
without.  Please don't make it the bulk of my code.

-- 
Barry Fishman