I was playing with the amarok collection for my school project and i found something strange while browsing ~/.kde/share/apps/amarok/albumcovers/cache/

Most covers are repeated under two names there - 100@somehash and 50@ or 100@artist_-_album
Which are safe to delete and if the hash-format is currently used then how to identify the files?

thx

you can pretty much delete every file in the cache dir. the original images are stored in the large/ folder.

the filenames are encoded for two reasons:

a) to avoid problems with file-naming conventions on several filesystems.
b) possible trouble with amazon licensing. \"you\'re not supposed to identify these cache files\".

have fun,
muesli

Well I just use Krusader so I can view the image, even encoded.
I think that protection via obfuscation is useless.
You can store amazon URL or the image on the HDD, but to obfusate the image name instead to use usefull and let the user decide how to be with the cover image.

//Cyrilo

i think this security by obscurity is useless
amarok has some function album --> hash
else it would not be able to display the cover
a reverse function may be dificult to find (see md5 etc)
which make it hart to calculate the album from the hash
but as one has all \"interesting\" album names
he could simply apply the first function to all
to find the correct file in O(n) with n:=number of albums stored

if amazon forces you to do so ok
but its trivial to get rid of

you dont get it, it has nothing to do with obfuscation. we store it as md5(artist + album) to get rid of all the i18n problems with various filesystems. we had problems storing umlauts on fat, storing certain special chars with xfs etc. etc. etc.

regards,
muesli

umm thats a point

i would have done other but hashing seems to be a good aproach to that too

btw who is insane enough to put his home on fat ?

thats another question anways... i didnt want to start changing the conventions with every other version, so i thought md5 is the safest way to go.

the otherway would be to let qt encode the filename. and yes, we tried that first, but it still failed for some reason.

regards,
muesli