Subject: Re: one more xdm problem
To: None <skippy@macro.Stanford.EDU>
From: maximum entropy <>
List: port-pmax
Date: 01/30/1999 02:10:10
>Date: Fri, 29 Jan 1999 21:21:24 -0800 (PST)
>From: Bill Studenmund <skippy@macro.Stanford.EDU>
>> Looking at the /etc/rc script, I see that xdm is not started
>> asynchronously.  If it so happens that xdm opens the console before
>> daemonizing itself, it's going to win the race every time.  If the
>> correct process is guaranteed to win the race every time, it isn't a
>> race condition.
>It may be very deterministic, but it's still a race. If something happens

I call it "synchronization".  The whole point of synchronization is to
ensure that the proper party wins a race.

>to xdm, then the getty fires up & you've lost the console.

If xdm crashes and you don't have getty waiting for the console, and
you're not on a network (and don't have other terminals), you have no
way to bring the machine down cleanly.  I'm not saying that it's a
good idea to have getty waiting for the console in all situations.  I
am saying that it is a good idea in some situations, and it does work
correctly in those situations.  So, your claim that it is necessary to
turn the console off was false.  Doing so may be desirable in certain
situations, but it is not required for the system to function in a
manner that could be described as correct.


entropy -- it's not just a good idea, it's the second law.