Subject: Re: Serial changes 1.1-current
To: Kent Vander Velden <graphix@iastate.edu>
From: David Gilbert <dgilbert@jaywon.pci.on.ca>
List: port-i386
Date: 05/02/1996 22:24:43
>>>>> "Kent" == Kent Vander Velden <graphix@iastate.edu> writes:

Kent> In message <199605020404.AAA13987@repeat.pci.on.ca>,
Kent> dgilbert@jaywon.pci.on.ca writes:
>> I recently had to revert back to 1.1 on my 386/40 box for various
>> reasons.  However, I'm finding the following curious problems with
>> my 16550's --- I can call with pppd just fine... and even run at
>> 38400 with my 28800 modem.  However, uucico fails with just a few
>> incoming bytes.  I cannot explain why.
>> 

Kent>   As I sent to the current mailing list a while back, with a
Kent> current kernel groups of characters are dropped left and right
Kent> nearly making the serial device unusable.  One other person said
Kent> they say similar problems but as to my knowledge nothing changed
Kent> to correct the problem.

	I used to have a current kernel (this is indeed a long story).
Right now the system is back to 1.1.  At current, I experienced some
lossage, but not enough to cause me constant worry... in short it
worked well at 19.2 w/ uucp and 38.4 with ppp on my 386/40.  For ppp
at 38.4, the 1.1 kernel is working better than the current.  For uucp
it seems to be not working at all.

Kent>   After having a day that was a little frustrating the following
Kent> may sound harsh.  I fail to understand why a "smart" serial card
Kent> would be required.  An OS would have to be very special to
Kent> survive if they force people to get special hardware.  I fail to
Kent> see the view of some people that seem to think that an OS should
Kent> not make up for the limitation of the hardware and instead push
Kent> the people into buying special hardware.  I intend to purchase a
Kent> PCI MB so I do not care about "that ISA problem" which NetBSD
Kent> will probably never overcome officially but a working serial
Kent> device is very important to us poor guys that lack a better
Kent> network interface.

	That may be.  However, it's also possible that 38.4 or 56K is
just too fast for a 386/40 all by itself.  I'd also like to run two
(or more) modems in the not-too-distant future.  I'm relatively
resigned to the fact that there is a limit to 'fast' serial chips.

	One other question, though.  The IDE board I have has fast
serial chips integrated --- they're in the IC, so you can't say much
about them.  However, another card that I have has a 16550AFN on it.
Is this the chip that has the working hardware handshaking in both
directions (which is not the normal case).

Dave.

-- 
----------------------------------------------------------------------------
|David Gilbert, PCI, Richmond Hill, Ontario.  | Two things can only be     |
|Mail:      dgilbert@jaywon.pci.on.ca         |  equal if and only if they |
|http://www.pci.on.ca/~dgilbert               |   are precisely opposite.  |
---------------------------------------------------------GLO----------------