war story: parallel(1) command

[Eric]

On Wed, 31 Jul 2013, Lennart Sorensen wrote:

> On Tue, Jul 30, 2013 at 06:33:44PM -0400, Eric B wrote:
> > Your "likely the same" is context dependent.
> > I agree with what you say above in the context of random file
> > corruption or in the case of files containing random bits.
> >
> > For Hugh's case, he wants to hash all the files in a real filesystem
> > to find real differences.
> >
> > If one calculates the SHA-N hash for each file, that would
> > answer the question ("Are these files the same or different?")
> > with virtual certainty.  There is NO need for an additional
> > compare if the same hash is found.
>
> Of course there is.  If you don't, you simply indicate you have no
> understanding of what a hash is.

Did you follow the link Hugh's provided?:
http://marc.info/?l=git&m=115678778717621&w=2

> > When probabilities are too astronomically unlikely,
> > they never happen in reality.
>
> That's not good enough for file comparison.
>
> Of course you are unlikely to find two different files with the same
> hash, so must likely the extra comparison won't happen on files that
> are not the same, but you still need to do it.

No, you do not need to do it, and doing it is a waste of time.
If something does not happen in reality, there is no need to test
for it.

In the cases I am talking about, think of a collision as being so
astronomically unlikely that it is indistinguishable from
impossible.
A distinction without a difference is not a real difference.
Any other factual event in this universe is more likely
to occur than such a collision.
For all intents and purposes, a real collision is impossible.


--
Eric B.
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists