downloading failed when PartsFile comes grater 2Gb

[alexrayne]

Hallow.

i try to download torrent that contains many files audiotrack and video that paired together. there more then 180 such pairs in a torrent. i download only audiotracks. piece size in one torrent = 16Mb. so every audiotrack have piese crosses on video at head and at tail. all amount of piececes = 180[tracks]*2[head+tail]*16[Mb/piece] = 5Gb

Vhen pieces file (~uTorrentPartFile_11BA46B55E.dat) comes over 2Gb excepted failure, and torrent stops downloading. and there is no vay to get it downloaded and start it to seeds.

[moogly]

Hum, I think there is another thread where a user met this issue too.

It sounds like a bug. http://forum.utorrent.com/viewtopic.php?id=64184

[alexrayne]

yep, this is the same bug. does it already fixed?

imho, thereis another way to fix huge partfile size - create torrent wich have no crossing pieces, or it is impossible? imho, if torrent describes mapping of files on a pieces, then it can map files, so that begining of a file starts at the beginig of piece, and the tail of file pads with dummy zeros to fill the piece. Why this idea not implemented?

[DreadWingKnight]

It is not implemented because of the backwards-compatibility breaking effect it has.

[alexrayne]

if succession is a major property, then maybe 2 versions of torrent (old + new) can be composed in one torrent file, so that old clients process only old torrent with solid mappling, and new can see optimised mapping.

if some motivation present, then some ways to solve any task can be found. anyway, somebody do such evolution in future, maybe not for utorrent.

[DreadWingKnight]

I'm telling you it breaks backwards compatibility because it has been tried by other clients and has been PROVEN to break it.

[alexrayne]

Can you give some links to see it closer?

and what about bug? does it fixed somewhere in betas or other clients?

[Twisten]

well, this is just an idea and might take some thinking to implement but how about instead of keeping the actual data in the part file, just keeping the hash result (or partial hash result) ?

[alexrayne]

>>well, this is just an idea and might take some thinking to implement but how about instead of keeping the actual >>data in the part file, just keeping the hash result (or partial hash result) ?

as i understand hash - is a result of function applyed to a whole datablock, it is not obvious - can it be stored as a function state on a part of block or it can be calculated on whole block only, it depends on function properties.

CRC function can store their state at any position of block and with this state calculation can be continued with any other data - least of block. but such hash function is very vulnerabe against hacking, thus such functions not applicable here.