I wrote:
> The silver satin flat modular line cord you are probably thinking of
> is not twisted.

"Paul B. Webster VK2BZC" <paulb@MIDCOAST.COM.AU> replied:
>   Except as a matter of how it's run.

I should clarify what I said.  Flat modular line cord does not consist of
individually twisted pairs, no matter how you run it.

I wrote:
> 10BaseT and 100BaseT cable has to be "straight-through", unless you're
> using a crossover cable between two nodes, in which case two pairs get
> swapped.

Paul replied:
>  As has been posted, the pairs are not adjacent.

I don't think I even implied that they were.  I deliberately didn't go into
detail on which pairs were used.

> They are calculated to be as near to two balanced pairs as can be
> practically implemented with ribbon cable.

No, they aren't.  They were chosen mostly arbitrarily.  The arrangement of
the twisted pairs within the outer jacket of the cable is not specified,
so there are no two pairs that can be known a priori to be better balanced.

Ribbon cable doesn't even enter into it.  You can't run 100BaseT over ribbon
cable.  But if that was a concern, the actual pair assignment used for
Ethernet would be suboptimal.  Using the 1/2 and 7/8 pairs would be better
than the standard 1/2 and 3/6 pairs.

ObPIC^H^H^HScenix:  If you don't mind overclocking the SX, it can run fast
enough to send Ethernet frames on a 10BaseT link in software.  However,
unless you operate the link as full duplex (typically on a port on an
Ethernet switch), you have to add external support for collision detection.
Unfortunately the Scenix can't be overclocked to a high enough rate to make
Ethernet reception practical without external hardware assist.

Cheers,
Eric