Questions from my local LUG ML

Conrad Knauer atheoi at
Sun Jan 7 20:49:28 GMT 2007

A follow-up comment and question from my LUG ML:

> I'm reassured in that a) they seem to be very conscious of these
> issues, and b) are very flexible in terms of what will eventually make
> up upstart 1.0. There's no guarantees, of course, but at least it
> looks like they have the right mindset in designing the software.

I would mostly concur with that assessment.  The one comment that
bothered me just a little was that "init and upstart are both userland
processes."  While technically true, PID 1 might be considered slightly
more important than most userland processes.  If inetd, or atd, or crond
hit a bug and fail badly, inetd, atd, or crond might need to be
restarted.  (which, i suppose, could be handled by upstart)  If init or
upstart, as PID 1, hit a bug that causes them to fail badly, the
consequences are somewhat more serious.

The concept of Upstart operating as the "process launcher" with some
sort of IPC between it and atd/crond or even inetd, makes sense.  I
would say that leaves it in the "do one thing well" camp.  Could shells
then also request upstart start a process for me, rather than forking it


More information about the upstart-devel mailing list