From: Nuno Silva (nuno.silva_at_vgertech.com)
Date: Sun 11 Aug 2002 - 01:51:45 BST
Hello!
That would be cpu and IO intensive, but would easy to compute md5
checksums of all files and then compare the ones with the same name and
the same md5 hash...
the real problem is: do you really want to do that to *all* files in the
filesystem?
it's ok to do that with /bin, /usr/bin, /lib and so on, but /home, /var
and /etc have files that can be changed by the system or the users.
examples:
----------------------
/var/log/syslog has one single line:
system starting up
/var/log/messages has one single line:
system starting up
do you really want to hardlink these files and loose some syslog output? ;)
or in /home you have /home/luser/test.c and /home/luser/backup/test.c
with the same content.
do you really want to hardlink these files and make changes made to
test.c hardlinked to backup/test.c?
one week after you'll be receiving support calls from the users to
restore some files - that were somehow "updated" - from backup :)
----------------------
With /bin the "unify trick" system works well because you don't edit the
file. the package management deletes the old bin/lib/whatever and
installs a new one.
YMMV of course.
Regards,
Nuno Silva
Peter Kwan Chan wrote:
> Well, yes. The question is how you figure out if two files are the same.
> (Computation-intensive).
>
> Peter
>
> -----Original Message-----
> From: ragnar_at_this.is [mailto:ragnar_at_this.is]
> Sent: Saturday, August 10, 2002 2:46 PM
> To: vserver_at_solucorp.qc.ca
> Subject: Re: [vserver] unify file system
>
> Hello agin,
>
>
>>The shotcut we're [ab]using here is that we know that:
>>
>> /foo/some/lib.so
>>
>>is much more /likely/ to have the same contents and length as:
>>
>> /bar/some/lib.so
>
>
> Just to confirm (could not find in the FAQ)
>
> On can not just unify all files found to be the same?
>
> ragnar
>
>