home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!munnari.oz.au!ariel.ucs.unimelb.EDU.AU!werple.apana.org.au!hal9000!monty!newton
- From: newton@monty.apana.org.au (Mark Newton)
- Newsgroups: comp.os.minix
- Distribution: world
- Subject: Re: minicom
- References: <1hre17INNcgl@werple.apana.org.au>
- Message-ID: <9301019220@monty.apana.org.au>
- Organization: APANA South Australia - State mail hub
- Date: Fri, 1 Jan 93 19:18:45 +1030
- Lines: 50
-
- hock@werple.apana.org.au (Warwick Hockley) writes:
- > At the moment I can't even get login to work properly with a modem. I have
- > to wait until the local modem has answered and hand shaking has taken place,
- > then kill login. Init starts a new login, and this one works - ie, gives a
- > login message. Everything is OK from there on. But you can't take this
- > seriously as a way to work.
-
- Actually, answering the phone isn't login's job; that task is supposed
- to be handled by getty. By the time login starts up, the line should have
- been ioctl()'ed to the right baud rate, etc.
-
- The problem is that the MINIX getty is woefully inadequate for the
- job (Sorry, Fred! I'm not criticizing you, just your getty :-). One
- would have thought that detecting the speed of an incoming modem
- connection was precisely the sort of thing that getty was supposed to
- do, but 1.5.10 getty hardly ever gets it right.
-
- I believe the problem is in the detection of line breaks. Sending
- a line break is easy (drop the baud rate to 300bps and send some nulls),
- but detecting an incoming line break is not really something that the
- MINIX kernel is very good at. I am still critical of the standard
- getty despite the fact that it's the kernel's fault because getty
- should have been designed to use something other than line breaks
- to detect the speed of incoming connections.
-
- Before I set up my machine for public access, I spent a day or so hacking
- /real/ speed detection stuff into getty. Now, if my getty is called with
- the -m flag, it sits on the line waiting for a CONNECT message. When it
- finds it, it just reads the baud rate of the new connection straight from
- the serial port.
-
- THIS WORKS, every time, without fail. The line break solution basically
- never worked.
-
- Now, why couldn't the original getty have been written like this, instead
- of depending on something as unreliable as line breaks for switching
- baud rates?
-
- BTW, I acknowledge that my Amiga hardware will have a different rs232
- driver to the PeeCees, and that the PeeCee rs232 driver might be better
- at catching breaks. If that is the case, then my criticism is undoubtably
- given from a 68000 bias, and should be read as such. Warwick's comments
- tend to suggest that the intel boxes share the same problems, though.
-
- - mark
-
- --------------------------------------------------------------------
- "Do not worry about an infinite newton@monty.apana.org.au
- supply of gibbons..." - Al Dearle Mark Newton
- --------------------------------------------------------------------
-