[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Mldonkey-tasks] [task #4755] BT: support for partial download

From: jave
Subject: [Mldonkey-tasks] [task #4755] BT: support for partial download
Date: Thu, 20 Apr 2006 13:57:28 +0000
User-agent: Emacs-w3m/1.4.53 w3m/0.5.1+cvs-1.946

Follow-up Comment #5, task #4755 (project mldonkey):

This 6th version of the prio patch supports:
- setting of priority on individual blocks
- setting of priorities on subfiles in a multifile torrent
- better view of subfile download progress

I renamed the commands:
debug_set_download_prio - for blocks
debug_get_download_prio - return block prio and download status
debug_set_subfile_prio  - operate on subfile level instead

Please read the help strings for the commands for details.

You must apply pango:s WIP3C patch before! WIP3C contains an important
bugfix in commonswarming, and the prio patch wont work without it.
(this is of course true only until WIP3C is applied to cvs)

the prio patch is furthermore tested with the "multiuser" patch, but
probably works withouth it as well.

there are some currently known problems:
- if you "merge" 2 downloads, the priority system will most probably

- block priorities doesnt survive core restarts

- a file is commited as soon as all non-zero prio blocks are
downloaded. This means that its possible for junk to be commited. You
are expected to know about this and remove junk files yourself. I'm
not sure I'm interested in "fixing" this, since there are no good
solutions in the general case.

Also, the implementation isnt the most efficient, but in practice it
doesnt seem to affect performance much for me at least. (there is a
linear block scan to determine if a block should be downloaded or not)

Furthermore the previous "download first" code, which hardcoded high
priorities for the first and last blocks of a file is removed by this

you can recreate the behaviour by setting higher prios for those
blocks with the commands.

Finally a warning: its not obvious to me that its good for the
networks and/or your average download speed to mess with block
priorities. Is it, for instance "fair" to download a single file of a
multifile torrent and then jump out?

In practice it appears to work well though.


Additional Item Attachment:

File name: prio_6.patch                   Size:11 KB



Reply to this item at:


  Message sent via/by Savannah

reply via email to

[Prev in Thread] Current Thread [Next in Thread]