"Forcing" gigabit link operation

- G
- googlegroups
  
  Contact options for registered users
Vote on answer
posted
18 years ago

Wed, Mar 22, 2006 5:38 PM

I would expect the 10BASE-2/5 jam signal to be even less relevant as a "message".

This thing was designed for a shared media with 2+ simultaneous talkers. There's no way that the talkers, nor anybody else could make any sense of what's going on in when many stations are shouting over one another.

In that case, the jam signal existed only to trip an over-voltage sensor in the sending transceivers.

The line voltage would be over the threshold because more than one system was using the wire at that particular instant. This is due to some analog voodoo which quickly goes over my head. :-)

In the case of UTP transceivers, send and receive are on different wires, so the jam signal *could* be meaningful, even if it isnt.

Cisco devices know all sorts of things about their link partners due to CDP. The definitely become aware of duplex mismatches this way. I have not ever seen collision counters being communicated over CDP, but I've never been looking for it.

/chris

- R
- Rick Jones
  
  Contact options for registered users
Vote on answer
posted
18 years ago

Wed, Mar 22, 2006 7:16 PM

Nah, then people would have been asking how to get their networks out of binding arbitration :)

rick jones

- R
- Rich Seifert
  
  Contact options for registered users
Vote on answer
posted
18 years ago

Thu, Mar 23, 2006 2:26 AM

I doubt that. The only Ethernet system that ever used a special "jam" signal was on early Xerox experimental equipment (i.e., Alto workstations). I doubt you used those (although it is entirely possible; I had an Alto on my desk in 1979).

By the time DIX Ethernet was released (essentially identical to

10BASE5), the special jam was gone.

- M
- Manfred Kwiatkowski
  
  Contact options for registered users
Vote on answer
posted
18 years ago

Thu, Mar 23, 2006 7:42 PM

A Longshine dual-speed hub negotiating 100-full to a Procurve 4000M.

As to the 10mbit issues:

3Com SuperStack II 1100 will negotiate 10-full but connected to some 3C905s will experience a duplex mismatch.

VIA-Rhine vs. ATI Centrecom 3600 hubs

Just from the top of my memory.

-- Manfred Kwiatkowski snipped-for-privacy@zrz.tu-berlin.de

- G
- glen herrmannsfeldt
  
  Contact options for registered users
Vote on answer
posted
18 years ago

Thu, Mar 23, 2006 8:02 PM

As I understand it, the 3C905, 3C905B and 3C905C are completely unrelated, other than the similarity in model number.

I once tried to get Solarix-x86 to work on a machine with a

3C905C, but it didn't notice at all. (In addition, I had a bad cable.)

I also used to have a 3C250, I believe on of the earlier models of fast ethernet (100 only) repeaters. I never had any problems with that one.

-- glen

- S
- Stephen Sprunk
  
  Contact options for registered users
Vote on answer
posted
18 years ago

Fri, Mar 24, 2006 2:26 AM

I don't know which model number(s) the problem was with, but I remember running a datacenter when they were popular, and I had this conversation several times per day, every day:

server admin: Your switch is broken! My server is coming up as [10/full |

100/half] instead of 100/full! me: Is it a 3com NIC? server admin: Yes me: Go get the latest Windows drivers. [ time passes ] server admin: It still doesn't work! You didn't fix anything! me: Did you reboot after you installed the drivers? server admin: Um, no. [ time passes ] server admin: It works now, nevermind.

If they were only repeaters, auto-duplex shouldn't have been an issue. I did see a few NICs that insisted on coming up at 10Mb/s when the switch supported 100Mb/s, though.

S