>From Scott Reynolds:

> The sgtty
> interface isn't really buggy in that it does what it was designed to do; 
> what it *wasn't* designed to do is what is causing the problems.  It's
> unfortunate that the code has to make an assumption of what
> characteristics the line has (because there just isn't enough info in
> struct sgttyb). 

Well, I don't know how the things work O:-) but how about this:

* user program calls the sgtty interface to set terminal properties
  * sgtty interface calls the termios interface to read the current
    terminal charasteristics
  * sgtty modifies the settings as requested by the user program
    (and touches only the ones that need to be changed, preserving the
    rest)
  * sgtty calls termios to se the TTY properties
  * sgtty returns to user program

I guess there's some reason why this can't be done, so go ahead and tell
me ;-)

No matter how easy porting applications may be, I still wish the existing
sgtty stuff worked.  Even if it forced the TTY to 8N1 would be fine..
I kind of think that fixing the sgtty stuff in kernel _once_ would be more
sensible than having 50% of NetBSD users fix 20% of the applications they
use by themselves ;-)

  -jm