tech-userlevel archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Proposal for write(1) addition


> > according to somebody from the #tmux-channel in freenode, tmux does not
> > set utmp entries because they would need additional permissions for that.
> That could be easily fixed by adding a utmp-writing daemon that took
> unix domain messages from clients.  Each message would pass both creds,
> so the user in question could be authenticated, and rights, so the
> utmp daemon could validate that in fact the client had an open file
> descriptor to the terminal it was claiming to be logged in on.
> This would have the added benefit that the utmp writing daemon could
> run as a _utmp role account instead of as root...
> Should work, no?
this would have been the cleanest solution, if you had it right from the
beginning on.
Adding it now would end up in complexity of another daemon running and using
memory for something elementary as write, being incompatible with all other
You would have to fix *all* programs (in pkgsrc) that create shells - that's
not only tmux, but afaik most of the X terminal emulations out there.

With respect to tmux, I don't know much about the import into base, but even
now it is not the most recent version (1.4 vs. 1.5), patching such things
would make it harder to update base's tmux.

This would be a *lot* of work for just a small issue (I don't know how long
the ptyfs issue was there but undiscovered)
I think it's just the way ptys are used that changed and it is nothing bad
(with screen starting login shells, `w` shows me >60 users on my system).
Instead of changing every single program into that raster, adapting the few
tools that read utmp is less work and more appropriate to the situation utmp
is used for.

Regards, Julian

Attachment: signature.asc
Description: PGP signature

Home | Main Index | Thread Index | Old Index