On 2006-06-13, Larry Qualig <lqualig@xxxxxxxxx> wrote:
>> If your architecture requires mandatory locks at the filesystem level,
>> I'd have to wonder about your architecture.
>
> One simple example is simply backing up your filesystem. Say you have a
> database file that's a couple of gigs in size. The backup app (or tar)
> needs to write a coherent version of this file to the backup media.
Of course, the database app has to be written to handle the case of
suddenly not being able to update its image at arbitrary intervals for
arbitrary periods of time. Hopefully it can at least queue them up for
later processing or something. Maybe it drops them on the floor...
Or you could write the database app to accept some signal to pause
updates while the backup proceeds.
Either way requires (nearly identical) changes to the app. The latter
doesn't require mandatory locks on the filesystem, though, AFICS.
> Because you may not have had to use this isn't the issue. The
> developers who created the OS wouldn't have bothered with implementing
> System V locking symantics if there wasn't a need for it.
They implemented *advisory* locks, yes. I was asking about *mandatory*
locks.
> Because the
> feature happens not to work correctly doesn't mean that it isn't an
> important feature to whoever needs it.
They work fine as advisory locks. I'll worry about mandatory locks
after I'm convinced I should care. :->
--
Sincerely,
Ray Ingles (313) 227-2317
Here, Iraq, take our Constitution. We're not using it anyway.
-- Robin Willams
|
|
