BG> I don't accept that the failure of ATZ to auto-baud the modem is a fault or 
BG> a bug at all.  In fact, I'm more inclined to suspect that it's been done 
BG> that way intentionally for some obscure reason (but yes, it should be 
BG> documented).

PE> It's fucked by design if done intentionally.

BG> Normal Hayes convention has ATZ resetting ALL registers to their previously 
BG> stored values, and USR appears to follow this convention to the letter, 

So which register has the speed then, Bill?  I couldn't see that
register listed.

BG> which is why the modem won't auto-baud, but reverts to the last port 

$600 and I don't even get auto-baud?  Crikey, one step above RPI.

BG> setting stored with &W.  Nothing wrong with that at all.

There IS something wrong with that, regardless of your ostrich
impersonations.

However, before we go any further, you are now trying to claim that
USR is following the Hayes convention.  Since Dave Hatch has a Hayes
modem, I just want to hear you say that if the USR works differently
from the USR, then the USR has a bug/design fault.  Then I'll ask
Dave to try it out on his Hayes.

BG> Can you give me just one valid reason for even needing to alter the port 
BG> speed on the fly? 

PE> What do you mean "on the fly"?  What happens is I decide to find
PE> out whether my system can cope with 115200, so I change my terminal
PE> speed to 115200 in my comms program. 

BG> That's what I mean by "on the fly".

Ok then, yes I do do that "on the fly".  I wanted my OS/2 system
to be running at 38400, but decided to load the SDL at 57600
because I was running under DOS and didn't really care if it
used up too many resources, as I wasn't doing anything else.
I wasn't happy to go to 115200 though, because it was my first
load.  However, if I was regularly doing it, I would indeed have
been using 115200, even though OS/2 doesn't support 115200 in
COM.SYS.

PE> I don't expect to go and have to issue a command to the USR to save to 
PE> NVRAM, that's what auto baud rate detect is all about.

BG> No, you try it first, and if it handles 115200 just fine, THEN you write it 
BG> to NVRAM with &W.  If one occasional extra AT command annoys you, I think 
BG> you're going to be awfully disappointed in life.

To try 115200, I have to run it for some time.  That includes
answering the "RING"s, which it never sees.  Yes, the bug annoys
me.  And I don't really care whether you think I will be 
disappointed in life or not.  Certainly I get great pleasure in
watching you scrape omellette off your face anyway.  BFN.  Paul.
@EOT:

---