> Actually, building a ``correct'' DB25<->RJ45 or db9<->rj45
> connector allows you to connect pretty much any RS232 device
> to any other RS232 device.  One can get away with this
> solution in more than 90% of cases (in my experience.  YMMV
> etc.) 
> Granted that a few signals (such as Ring Indicator) are lost
> but I've not needed them as much.

This solution works very well and has been used several 
places I've worked at.  You can get (they aren't common) 10 wire
RJ45 connectors.  What we did was have lines 2-9 (which correspond 
to 1-8 on an 8 wire RJ45) flipped as you suggest, but then 1 and 10 went
straight through and carried RI and CD (actually we only had one line
for ground and left the other open for a clock, we had a lot of sync
devices).  This still works for any to any since RI and CD are both
one offs w/o an "opposite" signal.  Compatibility was nice because
if you used an 8 wire line or adapter you just didn't get the
modem signals, such as when you plugged a terminal into a box that
supported a full modem port.

RI isn't very important, but having CD separate from DSR can be
important.

> Credit for this idea and many thanks for bringing some order
> into the chaotic world of RS232 goes to Dave Yost.

Mmm, the idea probably evolved independently in several places.
Apple used a variation in large scale on their Macs and Imagewriter II's
as early as the mid 80's.  Their serial cables (the min-DIN 8 ones)
were all flipped from end to end and every device had a DTE port.
So you could use one Apple serial cable to connect to an Imagewriter II,
another Mac, a modem (that had the magic plug\g, which apple's did, etc).
Of course they were RS-422 so they had to worry about +/- for TX and RX
as well.

In their tech literature, Apple credited the phone business which had
been doing it this way for years in many places.
--
Mark