UMSDOS and ZipSlack [LWN.net]

UMSDOS and ZipSlack

Posted Dec 8, 2003 18:48 UTC (Mon) by busterb (subscriber, #560)
Parent article: Microsoft aiming IBM-scale patent program at Linux? (Register)

Does anyone know if this patent would affect UMSDOS, a Linux FS that uses 8.3 FAT names and maps them to long, unix-compatible filenames along with unix file permissions using extra files to hold the meta-data. I know that ZipSlack and several other Linux-on-a-dos-partition distros use this method to install Linux without reparitioning. I don't know if UMSDOS uses any stuff from MS's long filename support.

(Log in to post comments)

UMSDOS and ZipSlack

Posted Dec 8, 2003 21:04 UTC (Mon) by proski (subscriber, #104) [Link]

/* 
 *  linux/fs/umsdos/inode.c 
 * 
 *      Written 1993 by Jacques Gelinas 
 *      Inspired from linux/fs/msdos/... by Werner Almesberger 
 */

VFAT first appeared in Windows 95 released two years later.

UMSDOS and ZipSlack

Posted Dec 9, 2003 10:07 UTC (Tue) by forthy (guest, #1525) [Link]

UMSDOS is either prior art or not affected by the patent, because it's implementation predates the filing of the patent. Other prior art of having long and short filenames together in one directory are the Rock Ridge extensions of the ISO 9660 filesystem (NM record for alternative name). The question still is if the way Microsoft does it is novel and different enough to qualify as innovation. IMHO Rock Ridge is closer to vfat than UMSDOS, which keeps the additional metadata in a file, not within the directory itself.

UMSDOS and ZipSlack

Posted Dec 9, 2003 12:27 UTC (Tue) by IkeTo (subscriber, #2122) [Link]

The idea of UMSDOS is to keep a separate file to tell which 8.3 filename "maps" to the long file name, and to store the uid, gid and umask which is also needed for a Unix-style directory. I don't think it can serve as a prior art of VFAT, which uses directory entries which are interpreted as "deleted" in the old filesystem implementation to hold the long filenames. The former would leave an extra file visible by the old 8.3 filesystem implementation, the latter would not. The former is generic and can be done to any filesystem, the latter is specific to FAT and won't apply to anything else.

I think the patent is probably valid (unless the objection that it is filed too late would hold some water). On the other hand, I don't think it really make sense for MS to ripe on it, since it would essentially kill the file format: there are dozens of better filesystem that (1) are free of patent issues, (2) have free implementations, and (3) have better performance than VFAT. Once MS start collecting patent money, vendors will start ditching the format (or fallback to 8.3).