Re: patch to avoid huge buffer in scan-skel.l

[Akim Demaille]

 >> From: Akim Demaille <address@hidden>
 >> Date: Sat, 01 Feb 2003 17:32:41 +0100

 >> The people I know who tried Guile had me convinced it is still too soon.

 Paul> Which version of Guile did they try, and what problems did they
 Paul> run into?  (I'm not a Guile user myself, but I keep meaning to
 Paul> give it a try....)

I don't know.  It was a bit more than one year ago.

 Paul> Please don't get me wrong: I like M4.  I'm just worried that it might
 Paul> not be the best tool for this job, as the job evolves.

I believe that the main difference bw Guile and M4 today is that one
has a maintainer :(

 >> I'm referring to everything that makes the syntax more pleasant (e.g.,
 >> named arguments), much faster (e.g., a non macro syntax for shifts),
 >> and an easy means to escape some characters (i.e., the "hard quotes"
 >> as he says).

 Paul> These are all nice things to have, particularly the last one;
 Paul> but at this point we are heading down the path of designing a
 Paul> language (since M5 is not what we want, either), and that
 Paul> sounds like a major investment.

I think there are many things we can have bit per bit.  Hard quotes
are nice, indeed, but quadrigraphs, thought quite a pain in the neck,
make it less urgent.  I'm especially impatient of getting the shifting
$@ stuff, as I'm quite convinced a lot of time is burned in m4_map:
currently it's n^2 :(

A simple Bison/Autoconf/M4sugar C module would already simplify a lot
a of things, and improve the speed.  m4_map and m4_quotearg are
scheduled to be the earliest additions...