Re: persistent ptys

[hpa@zytor.com (H. Peter Anvin)]

Followup to:  <4139F3FA.1070107@coppice.org>
By author:	Steve Underwood <steveu@coppice.org>
In newsgroup: linux.dev.kernel
>
> Hi,
> 
> It seems BSD style ptys are on the way out, and most systems will soon 
> have just Unix98 style ptys. This makes me want to move something to 
> Unix98 ptys, but I'm not sure of the appropriate way. The issue is that 
> things like HylaFAX expect to work with well known, persistent, names 
> for modem ports. A 100% soft modem in user space can easily provide that 
> with BSD ptys. With Unix98 ptys it is not so obvious what to do. Most 
> commercial soft modems don't have this issue, as they are part kernel 
> space/part user space designs. Obviously creating a link to a 
> dynamically generated pty with a well known name, and various other 
> things could be done. However, I assume other people have had to do 
> similar persistent pty things, and there is a well defined common 
> practice for it. Can anyone tell me what it is? :-\
> 

Two options: either continue to use BSD ptys (the kinds of stuff you
describe above is actually the one case where BSD ptys is a better
choice than Unix98 ptys) *or* create symlinks -- or if that doesn't
work for you, device nodes -- dynamically.

I.e. have a program which opens /dev/ptmx, gets a slave pty with a
specific ptsname(), then does a symlink() into a suitable directory.
The symlink is your persistent name.

	-hpa