On Wed, 25 May 2005 18:57:26 -0400, Olin Lathrop wrote:

> Byron A Jeff wrote:
>...<
> > Token passing is more flexible and also eliminates contention.
> 
> True, but it comes with its own set of problems.  If a node goes down and
> doesn't pass on the token things get messy, especially in a pure peer to
> peer environment.

True, but this was solved nearly thirty years ago by Datapoint in developing and implementing ARC (later 
"ARCnet"), when Ethernet and Cambridge Ring (the computer-, not spy- network :-) were still research projects.  
It is true peer-to-peer, with physical multiple-star layout but logical sequential token-passing.  It has a 
"who's there" process which means that each node learns who is either side of it, and it's triggered either by 
a new node appearing or an existing one disappearing.  There is no noticeable slowing down when this happens.  
We were using it in a live environment in the late seventies.  

I remember the day when someone mentioned that they couldn't "see" any of the equipment on the far side of the 
site (there was a canal running through the middle of the factory, with a bridge that carried the cables and 
pedestrians across) and it turned out someone had cut through a great wad of cables, having been told they 
were disused.  One of them, ours, wasn't!  The system carried on happily in two halves, and since the 
cross-canal requirement was only used a few times a day, nobody noticed for some time!  As soon as the culprit 
cable had been identified and re-made, it all reconfigured itself back as it had been.  It's a shame that 
Datapoint kept the technology proprietary for too long, giving the others a chance to get established and 
effectively shut it out.

>From another message on this topic:

> it is something I classify as "advanced"

When Olin says this, "Caveat Developer" !  :-)

Cheers,


Howard Winter
St.Albans, England


-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist