home *** CD-ROM | disk | FTP | other *** search
- From: srg@quick.COM (Spencer Garrett)
-
- I don't know if this is the proper forum in which to discuss this issue,
- but it seems to me that we're about to codify a poor implementation of
- the concept of symbolic links. IMHO symbolic links shouldn't have
- a format code at all, since they shouldn't be inodes. They make
- infinitely more sense when treated as directory artifacts - i.e. a
- directory entry either points to an inode or it supplies some more
- path information to be used in resolving the request. This eliminates
- all the silliness about what permission bits on symbolic link inodes
- mean, for instance. Here's the directory structure I use in my own OS.
-
-
-
- /* Directory entries are null-padded to a multiple of 4 bytes.
- * reclen and namlen fit in the first 4 bytes so that any amount
- * of free space can be described in a consistent manner.
- *
- * namlen == 0 => entry is free, reclen describes all adjacent free space
- * namlen != 0 => entry in use, reclen describes only this entry
- * ino == 0 => symbolic link (path follows name)
- * ino != 0 => file
- */
-
- struct direct {
- unsigned short reclen;
- unsigned short namlen;
- unsigned long ino;
- char name[DIRBLKSIZ-12];
- };
-
- Volume-Number: Volume 19, Number 26
-
-