OS costs

[Heiko Blume]

seanf at sco.COM (Sean Fagan) writes:
>You heard wrong.  So far, the FSF (not GNU) has not started development of
>an OS.  They *are* waiting for Mach (not MACH) to be freely available, but
>that depends on a few things (such as the DoD and/or the DoC, and CMU having
>it ready soon enough for them); if there are too many problems, they will go
>with another system, or, *worst case* write their own kernel.

the GNU Bulletin from June says a free version should be available this year.
if they can't get Mach they want to use MIT's TRIX or Berkley's Sprite.
(possibly Sprite will be it, then, the Bulletin says it's "at about the same
architectural level as BSD UNIX".)
btw: why should mach fall under any export restrictions ??? this kind
of protectionism makes me sick.

>>When this hits the streets with i386 and i486 patches to it, look out!!!
>>Free Unix with *SOURCE* here we come.

YEAH! the ratio of bug-fixes per month will be very interesting!

>Right.  First of all, Mach is not UNIX(tm).  It currently looks a lot like
>BSD, but that's because it also has Berkeley (and, therefore, AT&T) source
>code in it.  When that code is removed (a la Mach 3.0), the non-free code is
>gone, but so is the compatability.  One of the first things that the FSF
>would have to do with Mach is write a set of unix-compatable library
>routines.

they are currently *replacing* the at&t/bsd code, *not* removing.
code that can't be replaced fast enough will be placed in user 
processes that can be left out of the distribution.
the c library from GNU is said to be nearly completed.
they also want to use the sprite distributed file system, because
it's kind of hard to write a BSD fs from scratch...

>And it's *still* not going to binary-compatable.  Which means a *lot* for
>the '386 (and, of course, '486) world.  Nor will it run on a '286.  Nor will
>it run DOS code.

well, i don't have any facts about that. however, why should it be impossible
to run DOS stuff? as far as i understand it vp/ix loads DOS images, runs
them, and maps the i/o calls to something else, right? where's the problem?
if i can run apple ][ programs given the right hack, i can run DOS
programs given the right hack. (vp/ix *is* a hack :-)

apart from that, who cares? all those RISC architectures popping up are also
binary-compatible to nothing. that didn't stop the companies from selling these
machines. also binary-compatible doesn't mean the software works, right ?
('gee, this program expects to use /usr/spool/uucp/locks! my lock dir
is somewhere else!'.)

>More importantly:  nor will the FSF hand-hold customers, or put in a lot of
>work, so that non-wizardly people can install it.

so what? many people put in a lot of work to make FSF programs better,
in the gnu groups they just don't post bug reports, they post *fixes* too!
also, 'non-wizardly people' are mostly not able to install a full unix,
not to speak of adminstrating the thing. to get the hand-holding call
Cygnus Support (info at cygnus.com), they'll do it.

>>p.s. No wonder the Unix companies are trying to grab as much $$ as possible.

sure, but they'll never get as much people 'working' for them as the FSF,
not to speak of paying them.