Re: setting standard fds to console in translator libs

[Niels Möller]

tb@becket.net (Thomas Bushnell, BSG) writes:

> nisse@lysator.liu.se (Niels Möller) writes:
> 
> > Not even a write-only one?
> 
> Well, if you have access to the console, you can potentially tweak
> that terminal in various ways (like TIOCSTI).  We certainly don't want
> that.  

Perhaps it would help to be able to get a port to the console that
doesn't allow that (and then it should probably filter out tty control
characters etc as well).

> Eek!  So basically you are happy with all the users on a multi-user
> system storing their email in the syslog logs?  Because--trust
> me--that's what inevitably happens. :)

Disk space is cheap ;-) Seriously, if that is a problem, one would
want to (i) install quotas, and (ii) somehow set up per-user
translator logs. Is it possible to set up a "shadow-per-user
translator" that shows up as a single file-like node in the
filesystem, but redirects writes to different files depending on the
uid of its user? The user mux does something similar, but for
derectories, right?

> You're thinking of translators as things the sysadmin sets up.

No, I'm not thinking like that.

> But, we don't *have* a good solution to where passive translators
> send their error messages. [...] I'm just against forcing them onto
> /dev/console.

I'm just suggesting that sending diagnostic messages to stderr should
be a reasonable and supported thing for translators to do, and that
stderr should be set to some reasonable (and configurable) default by
the system, most likely the parent filesystem.

If you're concerned about users storing mail in syslog, syslog:ifying
all translators won't solve the problem.

/Niels