Re: A scary story from the real world.

[Brian May <bam@snoopy.apana.org.au>]

>>>>> "Per" == Per Sandberg <per.sandberg@bredband.net> writes:

    Per> One Microsoft person said "we had lots of crashes in the system and
    Per> the cause of that was that the driver vendors did not look on the
    Per> return code from functions"
    Per> Then the blamed the poor programmer for not reading the secret "users
    Per> manual".

I think some driver vendors could write buggy code even if a good
language was used. The general attitude in the Windows world is to try
and work around the problem as opposed to finding out why a driver
installation repeatedly crashes at a given point on a given computer.

I see only two solutions:

1. Microsoft write or review drivers themselves. Unlikely to
happen. Even with an open source model like Linux, some drivers that
end up in the kernel are horrible (or so I have heard); there are
simply too many drivers to review and possibly rewrite every one.

2. Move driver to a separate user space process somehow, so only that
one driver crashes instead of the whole computer. Also it should be
immediately obvious which driver crashed, so the complaints go to the
right place. (See "The Hurd" and "Minix" for examples).
-- 
Brian May <bam@snoopy.apana.org.au>