>    And thanks to everyone who replied. Although my problem still
>exists, I have learned along the way and have possibly narrowed the
>range of possibilities. However, I am still stumped.
>    I have <enormous snip>

Ok, you've tried everything,
--- so the problem must be something else :-)

Here are some off-the-wall ideas:

* Have you probed the cable end-to-end with an ohmmeter or
continuity checker in case the contacts have crimped badly?

* Have you looked into the NIC RJ45 sockets to make sure those
little contact springs are lined up correctly and not bent?
(You might check the continuity to the PCB traces too, in case
someone has given those secondhand cards a tough time.)

* Have you tried configuring both cards to TCP/IP and attempted
to PING one's IP number from the other machine? That should
bypass all that inscrutable Win95 networking stuff.

* Have you tried borrowing a pair of working cards and cable
from someone else (assuming you don't live in the wilds)?
When I had a Novell protocol problem on not-quite-PC-clone
workstations with missing documentation, I didn't solve it
until I got a pair of identical cards that had a diagnostic
which exchanged and counted packets: then I knew it wasn't
a hardware interconnect problem.

By the way, Bill Lee and Robert Prickett are arguing about
_the_same_ cable connections:

Bill has described the cable with the RJ45 connector
"Clip to bottom, Wire to right", whereas Robert uses
the official numbering, which corresponds to
"Clip to bottom, Wire to LEFT"
(numbering from top to bottom of view in both cases).

Since both Bill and Robert show both ends of the cable
this makes no difference to the connection routing.

The Hardware Book page HWB\co_Ethernet10BaseT.html has
a pretty picture of a RJ45 with Robert's numbering.

Len Warner <[log in to unmask]> WWW Pager http://wwp.mirabilis.com/10120933