My advice is to borrow from other clustered filesystems' experience. If you want to adhere to POSIX semantics then st_mtime and st_size visibility will be a particular headache, especially since you don't know when it's OK to lie (i.e., which callers of stat() are using st_mtime/st_size for synchronization).
Ideally we'd split stat() into metastat() and datastat()... Nico -- _______________________________________________ sqlite-users mailing list [email protected] http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
