🚀 go-pugleaf

In article <11771@amdcad.UUCP> phil@amdcad.UUCP (Phil Ngai) writes:
>nS/foot and there are 3 feet per meter. Then light takes 3000 nS to
>travel 1000 meters, 6000 nS to make the round trip. Ethernet heartbeat
>is supposed to be sent after the end of the transmission. If the
>transceiver returned heartbeat immediately after it perceived the end
>of transmission, the controller would not see heartbeat until 6 uS
>after it ended its transmission. Unless you "adapted" the transceiver,
>it would actually delay before asserting heartbeat.
>
>This is well outside the spec for both Ethernet 2 and IEEE 802.3.
>

This is not the only problem. The Ethernet 802.3 spec says that the
transmitting controller can expect to see its own data coming back on the
RD pair within 900nS of starting to send it out on the TD pair. If it does
not then it can assume that the transceiver has failed and cease transmitting.
I am not certain but I am told that the Intel LANCE depends on this
"sidetone" to work.

Mike

Disclaimer: Yep, I work for AMD and we make a LANCE too, but I'm a CAD
guy and really don't know anything at all about ethernet. My opinion
is not that of my employer.


-- 
-------------------------------------------------------------------------
Say NO to Spandex and chrome, say YES to silicon!

 UUCP: {ucbvax,decwrl,ihnp4,allegra}!amdcad!mike
 ARPA: amdcad!mike@decwrl.dec.com