Re: DiskDevice API v2.0

Tyler Dauwalder <tyler-cyCl/[email protected]> wrote:
> On 2003-04-05 at 08:20:14 [-0800], Ingo Weinhold wrote:
> [BDeviceList/notification stuff]
> > > > I believe, a `partition added' is more useful than a `device
> > > > changed'
> > > > event, even if the complete partition info isn't included in 
> > > > the
> > > > message (but only its ID). However, I think, it should be 
> > > > possible
> > > > to
> > > > improve the implementation, anyway.
> > > 
> > > Okay, then let's try to include everything needed to manually 
> > > update
> > > a
> > > BDiskDevice and it's children in response to a given notification
> > > without having to ask the kernel to do it for us.
> > 
> > I was first sceptical, because the amount of data to be sent, when 
> > the
> > hierarchy changes, might be (or rather: was) relatively great. But
> > since the granularity of the actions performed is rather fine now, 
> > this
> > shouldn't be a problem anymore. It becomes rather important, that 
> > all
> > notification messages are actually received, but by moving the 
> > message
> > delivery for notifications of all kind from the kernel into the
> > registrar, this should be rather safe to assume. As a fallback for 
> > very
> > rare situations the complete device can be updated (a change 
> > counter 
> > as
> > already implemented can be employed to detect skipped 
> > notifications).
> 
> I was wondering what the change counter thing was for, but I kept 
> forgetting to ask. I assume it's simply incremented each time it's 
> updated, so one can tell which version is newer by simple change 
> counter comparison?

Exactly.

CU, Ingo