Rodi - anonymous P2P

i know for sure that gimp and videolan and realalternative are under the fire. not in the IP court but they already remove parts of the sources from the distribution - read less codecs available for you, less effects, etc.
every single vendor of DVD copy software is in the IP court or on the way there.

I wonder if the uncrippled versions of these IP-infringing applications are available in an alternate site? Also, DVD-Shrink's development has now stopped (as terms of the lawsuit dictated), but let's hope that the code will be leaked into Opensource. I'd like to see opensource software exempt from all IP and patent law, but I know that will never happen.

for example, wouldn't it be nice to be able to stream data directly from DVD ?

And encode it to DivX in the process? That would be a great combination if a person could do all those things simultaneously. a current problem is that ripping a DVD and encoding it is a very time consuming process (several hours) - as is uploading it. But if those tasks could be done simultaneously, that might cut the total time in half. Application development could even be done legally (just use AnyDVD or some other 3rd-party add-on) as long as the developer did not instruct them to use a deCSS program.

the problem is that PCI bus throughput limits burst rate to ~150MBytes/s and sustainable write rates for average hard disk is in the area of 60Mbytes/s. it assumes that application writes consequitive blocks. If you to posistion head before wrting (seek time) you have a problem. in all BT clients developers make efforts to collect consequitive blocks before writing to the media.

I wonder if anyone has considered the possibility of arranging the BitTorrent file blocks by order of download sequence rather than dividing the original file download into blocks according to the actual file arrangement?

In this proposed example: when you start a download - nothing is written to disk until a block is completed, and these blocks would be arranged in the partfile sequentially in order of download completion. So a partial download might look like:

CCCCCPPPPP..................................
(C=complete block, P=partial block, ......=unfilled space)

instead of:

c...c....pcc....pp..........c..p.p.......c.......p ..

When all the blocks are complete, the "file" will of course be totally out-of-order, but a "file table" will keep track of the correct order of the blocks. I think that ultra-high-speed downloads might benefit enormously by this sort of arrangement, because the hard disk head would not need to jump all over the disk as a block completes, but could instead slowly move across the disk surface throughout the download. The physical order of the blocks on the disk - both complete and partial - could be changed ("defragmented") during the download process. (a highly fragmented HDD could also be utilized this way without excessive head-hopping because all the active blocks currently downloading would be grouped together).

Let's just hope that my above idea has not already been patented!

did you try to use BT with external drive ?

No, but I have tried eDonkey (original) with external HDD - this was the only way it would work with Goback.

yes, that;s the way typically clients use

But if those tasks could be done simultaneously, that might cut the total time in half.

i think that streaming directly from DVD is possible at low rates like 20K or 50K
ask Azureus team. i would like to hear what they have to say about it and i can not ask myself because they do not like me and my stupid questions very much

Originally Posted by larytet

there is another way
client writes blocks in the order they arrive. client keeps map of all blocks and rearranges them from time to time. eventually file will finish with blocks in the correct order.

That is what I was thinking. It seems that all swarming P2Ps use fixed block positions, so I agree that having a variable block position would help take the load off hard drives.

Originally Posted by larytet

ask Azureus team. i would like to hear what they have to say about it and i can not ask myself because they do not like me and my stupid questions very much

What? Maybe they don't appreciate innovative ideas? Is this the forum: http://sourceforge.net/forum/?group_id=84122 ?

I think that variable-position blocks will probably be a future feature in the next generation P2P applications, and could be done with current transfer protocols - so I'd be surprised that such a feature would not be accepted by current developers. Maybe they feel it would be too much work to develop.

Don't give up hope. More than a few times in the past I have made suggestions to P2P developers, only to have my arguments turned down repeatedly. Then, a few months or years later, those ideas would get implemented - sometimes by the same developers who originally rejected them.

i ment to ask Azureus team about streaming directly from DVD (i edited my post, sorry - it's getting rather late)

i undrestand that Azureus implements today the scheme you described, but i am not sure. Rodi today plays different game leveraging larger block size. i have RAM based caches both up and down. "downcache" i purge it from time to time, "upcache" works good because of the block size. eventually i will implement something along the lines of your post


this is interesting (unrelated) link
http://larytet.sourceforge.net/betaTest ... rsonal.php

if you are wondering what is it about read here http://www.integrityp2p.com/portal/foru ... stcount=12
and more here
http://www.integrityp2p.com/portal/foru ... stcount=32

P.S. this thread got 7K views.

...i did not give up on the Reed Solomon codes. i am learnign things. i need more time.

it would be nice to make Digital Fountain to donate their IP to the open source community. to make any real life tests i have to write code and my undrestanding is that any code i write should remain on my hard disk. i can not even upload it to the SF.net server. pretty stupid situation.


[Edit] ...and more i read less i understand how it can be used in decentralized systems

[Edit1]...multicast/broadcast applications which are (naturally/inevitably ?) centralized

As for the Azureus team including a CD/DVD-ripping feature in the official client, I can understand why they would probably not want to do this, as doing so might be begging for unwelcome attention. A much safer strategy would be to do nothing - and being open source, other outside developers would be able to release their own Azureus mod with this feature perhaps.

JiMiThInG reported that he burned out his CD-ROM by using it to share CDs directly on P2P, and this does not really surprise me, when a CDROM is operating 24h/d.

I would tend to doubt that Azureus or any other BT client uses a dynamic block map, though I can't say for certain. The reason I speculate this is that static blocks are so much easier to implement - perfecting the use of dynamic blocks would require a lot more code tweaking to optimize for all possible combinations of both fixed (user-specific) conditions (processor speed, HDD speed, free RAM memory, download speed, etc.) as well as variable conditions ( number of download sources, frequency of source disconnects, etc.)

I think that Dynamic blocks would probably be a poor choice for ED2K clients, at least as the network exists right now. One main reason is that eDonkey clients (using Horde enabled) are actually "traders" - eDonkey clients will upload for a few seconds, then disconnect if they do not receive a download in return - and do this over and over. As a result of eDonkey's Horde system, an eMule download will have a lot of partially filled blocks scattered throughout the downloading part-file's length. Dialup users also generate a lot of incomplete blocks when multisourcing due to preset time limits of each uploading client.

disk fragmentation is a huge problem and long "seek" time is another. i bet all clients implement this or that smart caching.


"streaming from CD"
i am not telling about sharing CD/DVD, but about streaming the content from the drive. the content is read only once.

hello,
i posted here http://p2pforums.com/viewtopic.php?t=25187

this is basically spam, but the project is very real and the server exists and allows to do everything you will find in the desciption. specifically access NATed and firewalled machines from any remote host. dynamic IPs, restricted NATs, etc. is not a problem if user can connect to port 80 of the proxy.

Proxy server is a top performance box and should stand rather large pressure. start to use it. it is really simple. it is simple to the extend that i use the proxy to copy files between Linux and Win32 PCs in my own LAN. it is definitely easier than yousendit.com and others.

see more http://gomyplace.com/install.html
or Source Forge http://larytet.sourceforge.net/goMyPlace.shtml