Thanks for Schwartz and Jay.I have known something from your words.

On 28 Nov 2005 08:48:28 -0800, [email protected] (Randal L. Schwartz) wrote:
>>>>>> "Jay" == Jay Savage <[EMAIL PROTECTED]> writes:
> 
> Jay> It depends on what you mean by "creation". stat() on POSIXish systems
> Jay> returns ctime, which is the inode creation time,
> 
> Nope.  Inode *change* time.  Will update any time you write to the
> file, or change any metadata (owner, permissions, number of links
> [including renaming], other timestamps).
> 
> Jay>  which is the time the
> Jay> physical location on disk was first written to (actually, these days
> Jay> it's the time the vnode was fist written to, but you don't really
> care
> Jay> that much about disk geometry, do you?). The time will not be
> Jay> preserved across relocations (i.e. mv and cp will destroy it),
> Jay> although many archivers reinstate the file's original ctime when
> files
> Jay> are restored from backup.
> 
> This is not possible at a user-level.  You can set atime and mtime from
> a system call, but not ctime.  ctime is always set to "now".
> 
> Jay>  For most applications, though, "when this
> Jay> particular disk space was reserved for data" is an effective synonym
> Jay> for "when this file was created. When you start moving files around,
> Jay> the questions merlyn raises become much more important.
> 
> Jay> For the record, Kirk McKusick--one of the original implemetors of
> Jay> UFS/FFS and the creator of UFS2--tells the story a little
> differently.
> Jay>  According to him, the FFS originally intended to use ctime as the
> Jay> creation time, but they discovered that dump needed a record of inode
> Jay> creation to make incrimental backups effectively. In fact, UFS2
> Jay> includes a btime (call it "birth time" if you want) field that is
> Jay> stable across inode relocations when the system uses btime aware file
> Jay> system operations. I don't know, though, if the Perl team has plans
> to
> Jay> update stat() to support it on systems where it is available.
> 
> Jay> Of course, that isn't completely at odds with merlyn's comments: the
> Jay> original unix developers started implementing file systems a decade
> Jay> before Berkely started work on FFS.
> 
> And, the whole notion of "creation time" doesn't make sense in the
> first place.  If I add a word, is it a new file, or an old file?  If I
> gut the entire contents, replace it with entirely new contents, is it
> new or old?  What if I rename it?  Is it new, or old?  There are no
> consistent answers to any of those questions, so why bother?
> 
> --
> Randal L. Schwartz - Stonehenge Consulting Services, Inc. - +1 503 777
> 0095
> <[email protected]> <URL:http://www.stonehenge.com/merlyn/>
> Perl/Unix/security consulting, Technical writing, Comedy, etc. etc.
> See PerlTraining.Stonehenge.com for onsite and open-enrollment Perl
> training!
> 
> --
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
> <http://learn.perl.org/> <http://learn.perl.org/first-response>


-- 
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
<http://learn.perl.org/> <http://learn.perl.org/first-response>


Reply via email to