On Fri, 23 Jan 2009 23:55:37 +0000, EAX wrote:
>I've downloaded from p2p the following archive:
>
>"Ls Land Issue 1 (Perfects) All Pics By Angelripper.rar"
>
>This archive contains 3017 pictures and completely matches bourne' csv
>except one file "lsp-014a-085.jpg", which is absent in the archive.
>I've compared bourne csv with mb/n@@dles csv and got the following
>results:
>
>1) lsp-013b-082.jpg has different crc. I've compared two copies of the
>file:
>
>>cmp -l lsp-013b-082.jpg lsp-013b-082c.jpg
> 7675 221 52
> 7676 13 166
> 7677 34 203
> 7678 204 31
>
>They have only four different bytes in the data segment! My check
>program shows both files authentic. jpeglib library decodes both files
>without any warnings. I think such corruptions cannot be found by any
>program. I've compared two files visually and found that mb/n@@dles
>copy is corrupt (color bar in the top of the image) while bourne copy
>is good.
>
>2) n@@dles writes in csv:
>
>"Set lsp-014a has 4 duplicate images. They have been included
>in this csv as this is how they appeared on the site.
>lsp-014a-007.jpg is a duplicate of lsp-014a-011.jpg.
>lsp-014a-008.jpg is a duplicate of lsp-014a-012.jpg.
>lsp-014a-009.jpg is a duplicate of lsp-014a-013.jpg.
>lsp-014a-010.jpg is a duplicate of lsp-014a-014.jpg."
>
>But Angelripper's archive contains lsp-014a-007.jpg, lsp-014a-008.jpg,
>lsp-014a-009.jpg and lsp-014a-010.jpg which are unique.
>
>3) lsp-014a-084.jpg is corrupt in bourne and good in n@@dles csv.
>
>4) lsp-014a-085.jpg has different crc. But as I already wrote
>Angelripper's archive doesn't have this file. So there is nothing to
>compare.
>
>The conclusion: any programs for checking JPEG segments and comparing
>files` structures with each other are generally useless. Official csv
>can be created only as a result of comparing two full site rips
>(including html and pictures/videos). If there are any differencies
>between them, one more unique site rip is needed to solve them. I
>don't see other way.
>
>Five files from Angelripper's archive on the way...
Great catch! It pays to have an extra set of eyes working on the
problem.
I went to my own archives on this. The first usenet posting I got of
issue 01 Perfects have file dates of January 2004. The four, so-called
duplicate files in set 14a were missing. The files jumped from 06 to
11. A repost of the set in April 2005 now had the gap filled with
duplicates of 11-14.
I've seen problems on LS sites with their html. Sometimes two links
point to the same spot. Could this have been the case and someone
renamed 11 to 07 because the 07 link pointed to 11? We'll never know.
I have also seen sharp-eyed guys spot such errors and manually change
the URL link to a correct address and recover files long missing.
Perhaps DarK-Elf did so. Angelripper, by the way, repackages others'
work, mostly DarK-ElF's, and adds a rip of the start site to the
package.
Or, perhaps the site changed at some point to correct the error. Take
your pick, as Naked Ape would say.
The corruption in lsp-013b-082 is very subtle. You are correct in that
no program can spot changes to the image. Only if control bytes get
corrupted will programs announce a problem. On my monitor, the
difference is that a minor color or brightness change takes place
starting about 6mm down from the top and subtly shades the rest of the
image.
Obviously bourne's lsp-014a-084 is horribly, horribly corrupt. But a
question remains about lsp-014a-085. This shows the uselessness of a
csv without files to go along. All of the copies of that image in my
archives are the same and match n@@dles.
I'll modify my own csv to reflect the five obviously good files even
though this will leave the three of us, n@@dles, bourne, and myself
totally out of agreement. Perhaps n@@dles will concur and join me in
the changes.
Again, great catch!
--
mr.bill
|
|