Re: [Mldonkey-users] alternate temp file system, plz

[Pierre Etchemaite]

Le Mon, 13 Oct 2003 16:51:14 +0200, Andre Hinrichs <address@hidden>
a écrit :

> 2) calculating checksums could be done when a chunk is finished or altered
> This could probably decrease the number of file derivates.

MLdonkey already checks chunk hash before sharing it (why would a new temp
file format be necessary for that ? MD4 routines can happily be run over
some offset range), and I guess other clients do the same.

> BTW, does anybody know why there are sometimes that many derivates of a
> file? I have seen some, where there are 40 and more!!! where there are
> some chunks with the same checksum and others not.

So we're left with data corruption happening later on (during the copy to
put chunks together for clients that don't use spare files for temps, during
the copy to commit files if temp and incoming dirs aren't on the same file
system, and last but not least, user's file duplication(s) -disk/disk,
disk/cd, cd/disk,...-)

Lots of personnal computers hardware is simply not good enough to avoid any
corruption over some Gb of copied data. Then sum up the probabilities for
all p2p users...