On Tue, 2009-01-27 at 15:24 -0500, Sowmini.Varadhan at Sun.COM wrote: > On (01/27/09 15:20), Sebastien Roy wrote: > > > > Consider the following three classes of administrative events on an > > object: creation, deletion, and modification. In the case of creation, > > it's easy enough to "undo", it's a simple deletion. The other two are > > harder. The log has to keep track of how to re-create an object that > > was deleted, or how to unmodify an object that was previously modified. > > I'm not sure I understand: if you were doing "delete -t", then you would > simply not remember the action in your persistent store, right? > Similarly for modify?
That is essentially what is being objected to; the implementation of temporary operations that have no effect on the persistent repository. The idea being proposed is that every operation results in modification of the persistent repository, but that the persistent repository that was modified is reverted to a previous known state. -Seb
