Chris,
> The OS is written to a much higher standard than most
> userland apps.
Kiddo, you just told me that even mere moving or copying could muck around
with the timestamp, so where are you taking that "higher standard" from ?
Also, you seem to be thinking that for "badly written software" changing
that "last written" timestamp is something that a program can "just do" -
without any particulary intent or reason. Newsflash: Its harder than you
think. :-)
> You just made it for me. Checksums are much better than last
> changed timestamps.
I would strongly suggest you read it again. I was making a case where doing
something "improperly" would most likely damage a lot more than just a
timestamp. In other words, the changed timestamp would than be the least
of your concerns.
> Checksums are much better than last changed timestamps.
Again trying to change the subject ?
> Whoosh! The OS and filesystem are not synonymous.
Just like checksums and hashes are not. You don't give a flying fuck about
it, but now trying to using distinction like that ? Really ?
And yes, I'm quite aware of it. But I mostly do not make a point of it
(especially not when its not relevant to the problem) as it just confuses
the issue. Just like I have not pointed out the difference between a
checksum and a hash.
Also, when was the last time you bought your OS and your filesystem
seperatily ? Lets guess ... Never ?
>> True, but now you're trying to change the subject to "what is the best
>> method to detect changes between two files", which I'm not going along
>> with.
>
> This was my original point to Shadow which you butted in on. Is there's
> one
> of us who's changing the subject, it's you.
[quote]
> I use a duplicate finder to check that the ones I have backed
> up to DVD are identical to the ones on disk. Good duplicate
> finders use hashes and checksums.
> []'s
You could cut out the middleman and store the checksums with your
backups.
The issue with using a duplicate finder down the line is that if it shows
they're different, you don't know which has changed.
[/quote]
This is what I responded to : Using a duplicate finder as part of a backup
process. Would you like to try again ?
> Nothing's perfect. However comparing timestamps is much worse
> than checksums.
You have said that several times, but no matter what I say you have not even
*tried* to come up with anything underbuilding it - other than a vague
hand-waving to "poorly written software", which I think I shot outof the
water.
> No. That is not what a timestamp is for. All it tells you is when it was
> last re-saved. You cannot make any judgement on state change.
True. The question is, if you have not changed anything, why would you
re-save ? And what if its intentional (the file is ment to be regarded as
the most recent one) ?
> For example I have a file with a timestamp of "01 April 2015 11:32:05".
> Has the file changed?
Im upping you one: How does your "lets take a checksum" indicate a change
(or not) ?
And when you figured that out, do you think I may also do a compare (of that
timestamp against another one) ?
>> If that timestamp has changed you may
>> assume that the file has changed,
>
> Nope. A file can be opened and re-saved without any changes
> occurring.
Yes, thats a possibility. But notice the "you may assume that" (for most,
if not all, intents and purposes). In other words, I'm quite aware of it.
Also, already replied to (repeating yourself doesn't make your case
stronger)
> Why is all this so hard for you to understand?
It isn't, and I've given several indications of that in my previous posts.
But I have the same question for you: With all my explanation, why don't you
understand that "just" using the "last written" timestamp normally works,
and when not it isn't destructive ?
Also, how is it that you seem to be acutily aware of how bad using a
last-written timestamp is, but have given no indication to the problems your
preferred method of using hashes has ? In short: You are not even
*trying* to compare the two.
And FYI, comparing hashes is just *one* step in the process. Which includes
generating them from full contents of the sourcefiles file (costs time). If
the compare is done to a backup hashes need to be generated for them too
{1}. Than comparing them and when they match *compare the actual file
contents* (costs time) to be sure they are really the same (and not just
hash collisions).
{1} If you are thinking about storing the hashes of the backupped files
somewhere that file can get lost, altered or corrupted. Or, even worse,
someone could alter a file on the backup which throws the list outof sync.
My suggested "compare 'last written' timestamps" (among a few things) do not
involve any of that ...
Bottom line: both methods have their pros and cons. The method involving
hashing will find *all* duplicates, but will cost a *lot* of time. The
"last written" method is fast, but could miss a few files which contents
have not actually changed.
But lets end this, shall we ? You seem to be convinced that I'm not
understanding any of it, as I'm convinced that its you (who doesn't even
try). :-)
Regards,
Rudy Wieser