In message <20001122125700.1D7D535DC5@smb.research.att.com>
	on Wed, 22 Nov 2000 07:56:58 -0500,
	"Steven M. Bellovin" <smb@research.att.com> wrote:
> >One simple way is the time when something created by open(), creat(),
> >mkdir(), mknod(), socket() and so on.  But I don't think
> >it is very useful.
> >
> 
> And what about link()?  That, I think, is the killer -- the times are 
It's not.  Although link() create another path name for a file, it
dosen't create inode nor contents for the file.  SInce symlink()
creates inode and contents for pointing another pathname, it is
creation of file.

> in the i-node, but "created" is a concept that applies more to 
> directory names.
UNIX's file don't have its name by itself, but directory hierarchy
gives the name.  So, link() is creating name not a file.

And Still I neve agree with adding support of creating time from now.

> That said, if memory serves correctly ctime *was* originally "created" 
> time for a brief period between 6th Edition and 7th Edition.  At least, 
> I recall seeing some references to that, way back when.  ctime first 
> appeared in 7th Edition.
Hmmm.  Since I've live from 4.2BSD, it's much impressive.

> It's also worth remembering that in those days, there was a lot of 
> pressure to keep the inode small.  Adding a 4th time field would have 
> taken away two disk block entries, which in turn would have lowered the 
> threshhold for when "indirect" blocks were needed, thus slowing access 
> to the file.  Features aren't free, and constrained resources often 
> produce far more elegant solutions than today's "memory and CPU are 
> free; why not add it?" attitudes.
I perfectly agree.

--
Takahiro Kambe <taca@sky.yamashina.kyoto.jp>