On Jan 20, 2008 9:25 AM, Richard L. Hamilton <[EMAIL PROTECTED]> wrote: > > Peter Memishian <[EMAIL PROTECTED]> wrote: > > > > > > > > > Anyone think maybe _FIOSATIME ioctl ought to be > > supported on more > > > > (ideally all) filesystems, and promoted from > > private to something more > > > > widely usable? Seems to me any really > > transparent backup utility would > > > > like to be able to do this, to not mess up atime > > as a consequence of > > > > backing up the file, nor ctime as a consequence > > of restoring the > > > > pre-backup atime. > > > > > > If we do decide to promote such a facility, I'd > > hope we'd find a less ugly > > > interface for it. > > > > What is is that you don't like? > > I suppose there are a lot of ways the objective could be achieved. > For example, an additional open() flag O_NOATIME to allow reading > a file without modifying its atime. That would have the advantage that > timestamps didn't need to be restored after-the-fact, so even an interrupted > backup utility using that flag wouldn't mess up any timestamps.
The only thing that bothers me about all of this is that bad applications will get this idea in their head that for "performance reasons" they should just use such options all the time. I know many folks turn off atime completely anyway because of the performance degradation (many GNU/Linux distributions certainly do). I do wonder how this will affect POSIX compliance. At last check, atime was required for POSIX compliance. So does this mean that applications using the proposed functionality from the start of this thread, or the functionality proposed above (such as a flag to open) would somehow violate tests? Probably not, but I'm hoping someone more knowledgeable can speak to this (Don?). -- Shawn Walker, Software and Systems Analyst http://binarycrusader.blogspot.com/ "To err is human -- and to blame it on a computer is even more so." - Robert Orben _______________________________________________ opensolaris-code mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
