Mark - Thanks for the papers. I have been working (at home...kind of off topic for the scsi group) on a paper about file system naming & integration. One suggestion has to do with a disk cache whose data lives on a faster disk (solid state, maybe, or regular vs. optical or jukebox, or just a faster platter). This sits underneath file structures and since disk is being used as the underlying store, the cache can be gigabytes large very easily; the cache uses the disk directly when this is done, and cluster coherency needs to be invoked much more rarely than with conventional memory cache techniques. Might be useful. The other has to do with having two sets of directories and mechanisms for doing it (without touching either VMS or file system code optionally!). Sounds like Spiralog, with its logstructured directories, would be a wonderful thing to use for the master level of this and the scheme might turn out to be able mightily to speed up read access even of conventional filestructures. (I'm presuming here that the 1 million cluster of ods-2 gets boosted...). I tried to mail it to myself here last night, but the mailer was wedged somewhere and it didn't get thru; I'll try again tonight. Perhaps I better send you a copy also. It isn't done...it's a think piece and I'm still working out some issues of alternatives that have to do with how files are searched for, and various points where algorithms might change, and some more. (I actually have the disk-on-disk cache thing designed and have a little code started, which I did around 3/1995; I've been busy with other things since and haven't completed it. BTW, I understand there is a Linux filesystem that can access NTFS disks available; the folks who wrote ntfsdos claimed it was their major source of information about what the NTFS on disk layout is. Do you have it or a pointer? Glenn Everhart star::everhart dtn 381 1497 (home: everhart@gce.com)