Tue, 27 Jun 2000 16:20:10 -0400 (EDT)
Steve Muir writes:
> This seems like an appropriate time to throw in my two penn'orth as one of
> the people who actually has written code for EROS.
That's worth about ten times as much as my tuppence (and Shawn's, etc.) :)
> First of all, enough of the basic libc-like functionality to allow porting
> of command-line type programs. I know some of this stuff is already
> present, like stdio and string, so perhaps we're already there, but a
> shell was still a long way off last time I checked.
I agree. I don't yet have a good enough grasp of EROS to have any
> Secondly, networking is obviously crucial. I see two big advantages to
> having networking up and running: server applications, which seem to me to
> be the most immediately interesting domain for EROS to attack, become
> possible; and if we can rlogin (and/or do remote X for those who really
> want to) then we don't need fancy local graphics capabilities, not even
> virtual terminals.
I am of the opinion that server applications alone are an evolutionary
dead end, but I agree that it is a direction in which making a useful
system is fairly easy.
> Thirdly, what about multiprocessor support? I know this is a very hard
> problem, but I claim that it's extremely important in the server
> appliance domain.
I think that eventually all computer systems will be multiprocessor;
I'm not sure that eventually all computers will be SMP. Supporting
non-SMP (and non-pseudo-SMP --- i.e. non-NUMA) computers requires only
kernel support for networking.
The vast majority of network servers are presently limited by available
bandwidth, not by CPU speed, and thus SMPing them gives no advantage.
<email@example.com> Kragen Sitaker <http://www.pobox.com/~kragen/>
The Internet stock bubble didn't burst on 1999-11-08. Hurrah!
The power didn't go out on 2000-01-01 either. :)