On Thu, Jan 24, 2002 at 07:11:25AM -0800, Scott G. Miller wrote: > On Wed, Jan 23, 2002 at 06:54:05PM -0800, Ian Clarke wrote: > > On Wed, Jan 23, 2002 at 08:58:54PM -0500, Gianni Johansson wrote: > > > Are you saying that this is fixed? > > > (Reported by Are) > > > http://hawk.freenetproject.org/pipermail/devl/2002-January/010122.html > > > (Confirmed by me) > > > http://hawk.freenetproject.org/pipermail/devl/2002-January/010129.html > > > > No, just saw it - it is clear that datastores can still be corrupted when > > the node is shut-down, this might be related to the fact that Freenet can > > only be shut down by killing it (although clearly in an ideal world it would > > be able to maintain datastore integrity regardless of when it is shut down). > > It may simply prove too tall an order for the datastore to be 100% > > resilliant to unexpected shutdown, I am not sure we should make this a > > pre-condition to 0.5. > Making a datastore resistant to corruption during shutdown is completely > trivial if it was designed correctly. The previous datastore, despite its > flaws at the application level, never had any trouble with this.
Actually, it did. We may never know for sure if the old DS corruptions were tied to node lockups, but if a file was deleted and its storage was reused, and the node then crashed before the next checkpoint, you'd wind up with trashed files. All my fault of course :) But it was solving that problem that led me to the new design. -tc _______________________________________________ Devl mailing list Devl at freenetproject.org http://lists.freenetproject.org/mailman/listinfo/devl
