Re: Improving API for lists, or wildcarded schemas

["Dave Cridland [Home]" <dave cridland net>]

On Fri Jan 30 21:29:57 2004, Tony Houghton wrote:
> In <20040130203415 9621 30044 polymer turner gestalt entity net>, you wrote:
> > Ah, so libgconf would obtain locks on gconfd, but this API wouldn't be
> > exposed. Still the same problem, I think - although most badly written
> > applications will be running through the library.
>
> If libgconf is written robustly it should be able to effectively
> guarantee not to crash while a key is locked.

While I have the greatest of respect for Havoc & co's programming skills, I 
think expecting libgconf to be *guaranteed* not to crash during a 
lock-get-set-release cycle is asking a bit much.

Gconfd, in any case, essentially needs to be written on the assumption that 
it will crash at the most awkward time possible.

> I was imagining it could lead to a loop, but now I think you're right.
> For one client to cause another client to have to reread the list, the
> first client would have to perform a write and therefore succeed, so a
> loop can't be sustained that way after all.

Oh, you can get loops from badly written apps, I'm sure. I'd hate to claim 
that a conditional store is a panacea, it's simply easier than having locking 
outside gconfd, in my opinion, and it's also more applicable to other tasks, 
like being able to append to a list safely from the app's perspective.

Dave.