Tom Haynes wrote: > Milan Jurik wrote: >> Hi Alan, >> >> Alan M Wright pÃÅ¡e v pá 30. 11. 2007 v 18:27 -0800: >>> I would recommend that, at a minimum, this proposal include >>> journaling. Some form of journaling or enhanced data >>> integrity guarantee is pretty much a requirement for file >>> systems now. >>> >>> If you include journaling during the initial implementation, >>> it will almost certainly be easier (technically) and lower >>> risk than adding it later. And it avoids a potential >>> on-disk format change later, which is always a good thing. >> >> ext3 as successor of ext2 didn't change on-disk format a lot, it is even >> possible to force mounting of ext3 as ext2. And future fork to ext3 is >> only depending on brain&manpower :-) >> >>> Is the plan to support the creation of ext file systems, >>> which would require utilities to do so, or just to mount >>> and work on existing ext file systems? >> >> As supervisor of the master thesis I requested only filesystem driver >> and mount utility. If somebody will need mkext2fs and fsck.ext2, he >> could recompile needed GNU utilities for it probably. >> >> Best regards, >> >> Milan > > Milan, > > ext2 is a non-starter for me. It would be worse than comparing a FAT16 to > a NTFS port. At least some people use FAT16. I don't know of anyone who > used ext2.
> So -1 as the project stands right now. > > I would prefer to see a project proposal which is to do ext3 and the > creation of ext file systems. > > It doesn't have to be the case that your student needs to do ext3 and/or > these file utilities. They can have the intermediate goal of getting ext2 > up and running. This is what I was trying to recommend. > And it may easily be the case that others will step up to do the remaining > work. Or you student may work on it after he graduates. In any event, > I much more prefer a forward looking project proposal than one with > a dead end. > > I'm also interested in how we verify that the implementation is clean-room > and > not tainted with a different license. Hmm, it is okay for the utilities to > be > GNU, since they are outside of the kernel. But we really need to make sure > that the file system code is clean. In Stephen's original implementation, ext3 file systems could be mounted as ext2 and vice versa provided they had been unmounted cleanly, which was nice for all sorts of reasons. This would suggest that having the proposal be ext3 wouldn't preclude some work being done later, however, this needs to be carefully thought out because one can't borrow from the existing implementation. Unless the extfs utilities have been cleaned up, there was a tight coupling between them and the file system. If this is still the case, the utilties would also have to be written from scratch. Alan > I believe that having ext3 available is an interoperability that we want. > It will > make it that much easier to have dual or triple boot systems. It will > allow > different OSes to share things like home directories on a laptop. > > Please modify the proposal to start us along the path of eventually > getting > ext3 in place. > > Thanks, > Tom > _______________________________________________ storage-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/storage-discuss
