Network problem

[Tom Fogal]

> I'm assuming Tom forgot to re-all on his note again so I've cc'd the group.

indeed, hahah.
i remembered this time though!

> Tom Fogal wrote:
> > 
> > Now that I look it up, it is not as specific as either of us thought. In 802.3
> > it is simply defined (in clause 1.4.81) as 'A condition that results from
> > concurrent transmissions from multiple data terminal equipment sources within
> > a single collision domain.' It doesnt mention frames or differences on transmit/
> > receive pairs..
> 
> Agreed, but the fact that multiple transmissions are taking place has to 
> be _detected_ somehow. The method implemented is to monitor the receive 
> pair while transmitting. If the pattern does not match, an error has 
> occurred.

It seems to be left undefined by 802.3; furthermore yes, it seems like
this would be one way for a vendor to implement CD.

> > CSMA/CD -may- be followed in full duplex mode, but it is not required.
> 
> Humm, the whole point of full duplex mode is that you can transmit while 
> receiving. All this really means is that its OK for the bit pattern on 
> the transmit and receive pair not to match. It is assumed that only two 
> layer two or higher system will occupy any given physical medium, thus 
> collision can not occur. Thus full duplex follows CSMA, but not CD.

hey! you cut that short. another couple sentences brings in that the 
lower (lowest, really) layers may still assert CSMA/CD but it is ignored
by the full duplex MAC. Practically, this means CSMA/CD is disabled on
full duplex links, but collision detection could still occur - as long
as the station doesn't act on it.

> Sorry to cut this short but I've got to bail,

ahh.. thats an acceptable excuse =).

-tom