On Wed, Mar 19, 2008 at 07:02:28PM +0100, Casper.Dik at Sun.COM wrote: > > > >Boy, y'all misunderstood what this proposal was about: > > > > _augment_, not replace. > > *sigh* > > > >I've tapped a raw vein of dislike of SMF, a blinding dislike. > > It was not my intention to start a flamewar; not should you infer > that I dislike SMF. It has, however, some rough edges that I think
Understood. I'm not accusing anyone wanting to start a flamewar, nor do I think *you* have a blinding dislike of SMF, but that some participants may well have such a dislike (enough dislike has been expressed so far). > we need to address before we continue blindly where we seem to > be going now. I agree that it has rough edges. > >(Though I'm quite interested in the UFS/SMF/SQLite reliability thread. > >I blame UFS though. ZFS boot, we need you.) > > I don't think that it is justified to blame UFS; if you build on top of > quicksand without proper foundation, it is YOUR fault the building > collapses. > > It is sometimes difficult to understand the intricacies of doing UFS I/O > safely but when you create a database engine you MUST take due care > and attention (and if you use that database engine, you MUST check that > it works properly). I'm pretty sure that the SQLite community really cares about this. And I think we need much more info before we blame any one component (so I blamed UFS too soon, sue me :) My recent experience with UFS left me thinking it's just not at all reliable in some circumstances. > UFS as a root filesystem at the time of building SMF was a given; so it > must work and be reliable in that context. Has UFS reliability degraded since?
