Re: [Nmh-workers] Locking (specifically, sequences)

[David Levine]

Ken wrote:

> >> AFAICT, if you reread the sequences file, that will solve that
> >> problem.  I cannot think of a scenario when it does not; can
> >> you think of one that doing that will fail?
> >
> >Not off hand, but I can't get beyond "merge conflict".
> 
> Maybe you're thinking of it wrong.
> 
> We have message N.  For each sequence, there is 1 bit of
> information:  is it on this sequence?

I think that answers it.  We're limited in what we can
do with that 1 bit, and the bits are independent.

> Now think about nmh operations: do we ever get into a
> situation where one command wants to add a message to a
> sequence, and another command wants to _remove_ the same
> message from the same sequence?  That's really the only
> possible "merge conflict" we're talking about here.

Even that doesn't bother me.  All but the last command will
lose the race, which is what we always want.

David