Jerry Coffin was saying to Tika Carr on 04-15-97  09:32:
 TC> 2. Remember that a 486 is not a "true 32-bit" machine. I was told that
 TC>    only a Pentium (ie. 586 and 686) have true 32-bit data transfers.
 JC> This is just plain wrong.  486es have a 32 bit path to memory.
 JC> Pentiums (and above) have a 64 bit path to memory.  Both use 32 bit
 JC> motherboards.
A friend who builds computers told me that all 486s were 16-bit and I
too somehow feel she may be wrong. I could SWEAR my 486 is true
32-bit! :) But I went on that I guess thinking maybe she's right...
 JC> The real way to diagnose the problem is for Danial to post his code so
 JC> people can look at it and test it out.  My guess is that some bug has
 JC> lead to things being written to the mailslot twice, but it's also
 JC> possible that this is simply a bug in Win95.  One way of testing that
 JC> would be to see if Windows NT does the same thing or not.
Hmmm... I have Win32s, but not Win95 (yet). I was thinking it was
because the throughput (ie. 16/32/64-bit) was different from one
machine to another. When it sends stuff out, the bus on a 486 (going
by your explaination, which I'm more likely to believe) is only 32-bit
where the Pentium PCI bus is 64-bit (here I go again theorizing...)
but if this is true, then the 486 is only sending a 32-bit stream of
data to the 64-bit Pentium, which wouldn't the Pentium get the data in
two 32-bit chunks? So he'd basically have to get only one of the
32-bit chunks each time a transmission is made...? I also wonder why
the Pentium wasn't taking in 32-bits of data and another 32-bits of
garbage tagged on to it to fill the other 64-bits.
And I have had "repeat data" problems in my (non-network) programs,
when I didn't use pointers right.  ;)
Tika
... "Dave, put down those Windows disks. Dave, please. DAVE!" - HAL 9000
___ Blue Wave/DOS v2.30
--- QScan/PCB v1.17b / 01-0406
---------------