13 May
2002
13 May
'02
7:37 p.m.
On May 13, 13:58, Mark Tapley wrote:
I think they do; the one I built (matching
instructions given
previously) does, with a caveat: I've used it to connect two (OT, sorry!)
Mac Powerbook 3400's. The two machines in question have to have Appletalk
changed to use ethernet within a few seconds of one another - otherwise
they each assume they are on a "dead" network and will not recognize the
other machine when it comes up. This may be a unique feature of the
Powerbooks. It is *not* a problem when the two machines are successively
hooked into a hub - even minutes apart - so I don't think it's a problem
with the Appletalk implementation, but I don't know.
Anyone else with more experience on this? Do hubs do any
keep-aliving for their ports, or recheck periodically for connections? Do
other two-node networks need "pinging" from both ends simultaneously to
bring up?
A hub -- or indeed any other 10baseT or 100baseT network device -- is
supposed to emit a stream of "link pulses" all the time it's available for
use but not sending data. The link pulses are short but regular (100ns at
16ms intervals) and only positive-going, so they're not confused with data.
These are the same pulses used by 100baseT and 10/100 devices for
autonegotiation.
In the case of the Powerbooks, what's happening is, as you say, that the
software is deciding that the network is dead and shutting down the
interface. Yes, that's almost unique, and certainly not the normal way to
operate. I've come across only two other devices like that, and in both
cases they were devices that fell back to 10base2 if they didn't see link
pulses within a very short time after initialisation.
--
Pete Peter Turnbull
Network Manager
University of York