Has there been any thought given to the idea of using the ELF comparison at publish time so that when a file is published with the same ELF content as a file that is already in the repository, no new file would be created in the repository, and the action would be set so that the hashval is the same as the existing file? This would eliminate the need for having ELF hashes in actions and would still get the benefit of reducing downloads for "unchanged" ELF files, while at the same time eliminating this problem of having files on different systems that are both the "same" and "different" (depending on how the files are compared).

Tom


On 01/13/10 07:08 PM, [email protected] wrote:
On Wed, Jan 13, 2010 at 11:58:26PM +0000, Chris Gerhard wrote:
On 13/01/2010 23:40, Shawn Walker wrote:
That's what pkg verify is for.
I don't understand. I certainly did not trust pkg verify over a cmp
of the files and would not expect anyone too unelss it is documented
in very big letters.
It's up to you to decide which software you trust.  However, the verify
command is working correctly and the sections of the ELF file that
affect its behavior haven't changed.  If you don't trust us, fine, but
don't blame us for using the wrong tools.

If the system's security had been compromised pkg verify would
report no problem as the hacker would make sure of it.
As Shawn has already observed, if the system is compromised you can't
trust any of the software on it.  The cmp and digest commands could have
been modified -- or the C compiler for all you know.  See Ken Thompson's
paper "Reflections On Trusting Trust" if you really want to get paranoid.

http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.84.8238

Yet the binaries are different. How this is not going to lead to
confusion or worse I can't imagine.
Again, the binaries aren't different as far as pkg is concerned.  While
a portion of the file may have changed so that the digest of the entire
file is different, the ELF sections of the file that control its
behavior remain the same.

-j
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

<<attachment: Tom_Mueller.vcf>>

_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to