to comp.lang.forth

Francois-Rene Rideau rideau@clipper
Wed, 4 Jan 95 19:15:38 +0100


In-reply-to: st000533@hrz.th-darmstadt.de's message of 4 Jan 1995 01:45:44 GMT
Newsgroups: comp.lang.forth
Subject: Re: File system as FORTH database app (was Re: Operating Systems as a Needless Complication)
References: <3e16qi$6ho@life.ai.mit.edu> <40.3805.1617@channel1.com>
	<3eaenl$1e22@rs18.hrz.th-darmstadt.de>
	<3ecuk8$1443@rs18.hrz.th-darmstadt.de>
Distribution: world
Mail body next line:

In article <3ecuk8$1443@rs18.hrz.th-darmstadt.de> st000533@hrz.th-darmstadt.de (Wolf Lichtenberger) writes:
   How about creating at least the file system part of an OS in form of
   a database application written in FORTH ? Actually, i don't care so 
   much for it being in FORTH (were it not for this '..faster, smaller and 
   better..' above), as long as it's both (DB & FS), with the capabilities
   of being both. Methinks i've heard of some such beasts (like MUMPS, 
   maybe?), but they've IMHO never penetrated the market segments we are
   aware of.
   'Nuff said. Comments please ?

   A new trend in OS research is for persistent systems: the concept of files
does not exist anymore; instead, objects persist until they are not needed
anymore, as on databases. The difference is that *any* object (including
functions/programs) can survive this way. Thus the user doesn't have to
care about loading variables from disk, checking file consistency, building
complex error recovery if such things fail, etc.
   See the fine work of University of St. Andrews (Scotland) about persistent
systems. A database is just a persistent object among others: some associative
array, set, list, of structured elements.
   Some FORTH or FORTH-like systems (like the holon FORTH system, or HP
calculators) already use such technique. But plain ANSI FORTH is too
low-level to allow both efficiency, consistency, and liveliness in such
systems: persistence means that no program should crash the system or be
a resource hole; it means that there should be some garbage collection and
data consistency mechanisms that every program will respect.


   I myself am intending to write a persistent (and distributed) operating
system over a forth-like kernel, and, if not an ANSI FORTH system, providing
the great concepts of FORTH, dynamism and reflexivity with direct user
interactivity, while bringing a garbage collector, and thus a strong but
dynamic type system.
   The system is still in design phase, though the general goals seem to
be settled; project organization is still not perfect, and some important
points are to be decided.
   Implementation phase has not begun; but currently only a i386 PC version
and a POSIX-based C version have found voluntary implementors; implementors
for other platforms are welcome.

   For those interested, join the Tunes mailing list:
mail -s "foobar" listserv@ens.fr <<END
help
sub tunes my name
END
   The subscribee is the mail return address.

   The mailing list archive are available by ftp at
frmap711.mathp7.jussieu.fr:pub/scratch/rideau/moose/