[unios] Re: Networked GUI (was Posix and Networks)

[Pieter Dumon]

From: Pieter Dumon <Pieter.Dumon@rug.ac.be>

> 
> From: Pat Wendorf <beholder@ican.net>
> 
> >                  *******************************
> > GOD ! What are y ou talking about? Locking out the supervisor terminal?
> > OF COURSE. Ever been a sysadm???? Do you know what it feels like being
> > root on a multi-server network??? Do you know what one would do to obtain
> > this? Of course a supervisor terminal cannot be reached. It's the most
> > important thing in the whole system! Such as simple security is something
> > that is VERY BASIC to implement. (At least on servers)
> > Also,we really don't want to make the networking errors Unix made back in
> > the seventies and NT does now again. NT doesn't even survive a
> > denial-of-service attack! You have to add third-party software.
> >                  ******************************
> 
> Wow strong feelings :)  I believe we stated it just for the sake of stating it :)
> It would not have been forgotten.

Sorry about that. I spilled my coffee while reading it and almost ruined
my keyboard. I was a bit pissed, I guess...
But still...

> > > I have another suggestion. Redundance in networking, which allows one or
> > > several nodes to fail without loss of the joint computation or information.
> > > To how big degree this can be done by the OS I don't know, but it's
> > > certainly useful. What if you could just declare some data to logically
> > > exist on the network, and it's automatically mirrored on several nodes,
> > > with accesses made to the nearest copy without any effort from the
> > > application's side?
> > > binEng
> > >
> >
> > Hm. What you are referring to, is ont of the buzz-words of the last year,
> >  CLUSTERING. Unix can do it, MS just starts to get some in-the-field
> > experience. It's really important for a corporate server-OS.
> >
> > An OS that can run _really_ non-stop (off course on the best hardware),
> > should be able to be implemented too. Although... Tandy and some others
> > seem to be the only firms able to do this... So it is rather difficult, I
> > think.
> 
> That's kinda what I figured, but with our model it might be a little easier to
> make.  We could have the primary machine make memory images of itself and transfer
> them to the backup, and when the primary goes down (stops sending, or is observed to
> be in lock-up), the secondary can jump into action at the same point the other left
> off...
>
> Only one problem I see here, and that is having a GIG of working system in memory,
> and getting that to another machine... FDDI or ATM I think would have to do it...
> 

It's LOTS more difficult than that. Even with a good design I don't know
much about clustering, and there aren't many people who do know.


Pieter

----------------------------------------
 Pieter.Dumon@rug.ac.be               
                                      
 http://studwww.rug.ac.be/~pdumon     
 
 ICQ  : 12428974
---------------------------------------

------------------------------------------------------------------------
To unsubscribe from this mailing list, or to change your subscription
to digest, go to the ONElist web site, at http://www.onelist.com and
select the User Center link from the menu bar on the left.
------------------------------------------------------------------------
UniOS Group
http://members.xoom.com/unios