[Resolvconf-devel] Bug#213591: Serializing file access

[Thomas Hood]

Just now I wrote:
> To be correct, though, we would still need to serialize
> resolvconf's accesses to individual records, and watershed
> wouldn't help with that.


On second thought, it *isn't* necessary to serialize resolvconf's
accesses to individual records.  Resolvconf only either overwrites or
deletes records atomically; it never reads-modifies-writes records.
I.e, it's idempotent.  So no record corruption can arise from two
resolvconf processes running simultaneously.

(The end state of the database does depend on which resolvconf process
acts last, but this is not resolvconf's problem.)

So the worst thing that can happen is overlapping or redundant update
runs --- misbehaviors that would be eliminated by using watershed.
-- 
Thomas