Ingo Weinhold <ingo_weinhold@xxxxxx> wrote: > > Actually we do that with attributes in general. I guess it could be > > extended to do it with files too but I am not sure if we would > > really > > gain much with it as files would have to be pretty small to fit in > > what > > today is the small data section, Compression would help here. > For a new FS I'd rather go the ReiserFS way and get rid of the inode > notion > completely. That this is a worthwhile endeavor can be easily seen > under > Linux by letting ReiserFS 3.6 compete e.g. against ext3 on a file set > with > many rather small files, say a Haiku tree checkout. Definitely, and even Reiser 3 has much room for improvements (I guess that's what Reiser 4 is about ;-)). > AFAIK for Reiser 4 it was intended to support compression. Don't > know, if > that was actually realized, or whether any benchmarks suggested that > this > is a good idea. Does probably heavily depend on the transfer speed of > the > underlying storage device. I would suggest adding a generic compression layer, which would enable the feature for all file systems; for those that support attributes, the compression could also be enabled on a per file basis. Bye, Axel.