-=> /* Quoting Lynn Nash to Frank Sexton */ <=-
 FS> How was IBM able to build more than one machine the
 FS> same if there were no specs?  I thought they had XT
 FS> specs and AT specs that they went by and that these
 FS> were eventually adopted by "clone" makers.
 LN> He is right, there were official XT spec's but there
 LN> were no official AT specs.
FS>Perhaps I should have used the phrase "accepted AT buss
FS>standards" instead of official.  The history lesson was
FS>interesting.  So, IBM thought the AT would last about one
FS>year eh... (grin).
FS>My only point was that virtually everyone (software and
FS>hardware folks) conform to the COM1 and COM2 address (and
FS>IRQ) standards (de facto or not).  But that many video card
FS>manufacturers do not feel the need to adjust their
FS>addressing requirements so as not to interfere with the
FS>operation of COM4.  This can sometimes cause trouble for
FS>COM4 users with certain video cards.
OK, I missed this part, and thought it was common knowledge, at least
in heavy tech circles.  This is not the fault of the video card
manufacturers and it occurs purely by accident.  The VGA control
register is at I/O port address 46E8H while COM 4 is generally at I/O
port address 02E8H.  Where the conflict occurs is that COM ports are
generally considered to be 8 bit devices with designs from way back to
the original PC.  As such many systems or COM cards only decode 10 bits
of I/O port address; all the original PC needed.  The video card is 16
bits and when it is happly being addressed as 46E8H it looks like 02E8H
to the COM 4 port which does not see the top 6 address bits.  The video
card is in no way invading the COM port space.  It is not the video
cards problem and goes back to the lack of an original official AT bus
standard.  Remember my comment about things falling through the cracks
in the last message?
Here is a perfect example but people tend to blame the VGA card makers
for following the VGA standard.  This is because Mono, CGA, and EGA did
not have this problem since they do not use address 46E8H. The
folklore about renegade video makers comes from old time users
(generally BBS sysops) of 4 COM port systems.  This is a generalized
hardware problem pure and simple which no one is willing to take the
blame for.  If your machine decodes all 16 address bits, used by 16 bit
I/O cards, for its COM ports then the problem does not exist.
Another that raises its head sometimes is the fact that memory on the
ISA bus must be accessed in 128K increments. A mix of 8 and 16 bit
cards that use close menory addresses will fail. This is because the 16
bit access signal MEMCS16 is raised by 16 bit cards for any memory
request within its 128KB addressability block. If the memory cycle was
intended for an 8 bit card, memory addressability falls apart. Video is
affected most but you can run across this with 8 bit buffered Lan cards
with addresses close to the video buffer region. Arcnet always had
problems.
Also mixing 8 bit mono cards with VGA in an attempt to have dual
monitors causes performance loss.  Newer VGA cards would drop to 8 bit
operation when this happens to prevent conflicts.  People would then
complain about lousy VGA video performance when the card was just
protecting them from themselves.  The real problem was that since mono
cards had the primary printer port people would stick them in just to
get a cheap printer port not realizing the performance implications.
All in all it is really too bad that market forces let the AT bus
escape, from the lab and mutate into ISA, at about the time the PC
really started to take off.  It would have been better to wait an have
gone straight from XT to MCA.  Unfortunately IBM also had other ideas
about that and Compaq plus others were already exerting market pressure.
--Lynn
 * SLMR 2.1a * We are OS2 developers; wierd is part of the job.-Janeway
--- DB 1.39/004485
---------------