On Wed, 10 Oct 2001, Patrick Oscar Boykin wrote: > On Wed, Oct 10, 2001 at 12:19:46AM -0500, thelema wrote: > > I'd like to clarify my position, which probably seems to be against > > redundancy if you just read the above, but I really think that we should > > try non-redundant splitfiles and if they really can't be requested > > successfully, we should try adding a *little* bit of redundancy and see > > how that improves things, and not to go overboard with the redundancy. > > This is a well studied problem (percolation, erasure channels, channel > coding, etc...). There is no reason to blindly hope this will just work > out. Look at the mathematical results. At the same time, I agree that > you should not "go overboard with the redundancy". Someone should look > up an appropriate code to use here and apply it. This should not be > done by the seat of one's pants. The great thing about Shannon's > capacity result is that you don't need to go overboard with redundancy > in order to get exponentially close to perfect transmissions. > There has been a lot of work on error-correcting and erasure codes. The problem is their model of transmission involves a simple channel where each bit (or symbol) has an independent chance of being recieved as erased or as corrupted. Freenet is much more complicated in that the chances for recieving each piece is a (very complex) function of how much that block has been requested and how much other blocks have been requested. Because of this, the standard analysis doesn't really apply. This is part of my argument against saying "there's an independent [90|99]% chance of a piece being retrieved." That's not a very realistic model of what's going on, and I think it's too pessimistic.
> Here is some code that does a lot of various schemes. A code should be > selected that protects against burst erasures (like a splitfile not > showing up): (this page might not have a good one for that, I didn't > read carefully) > > http://www.csl.sony.co.jp/person/morelos/ecc/codes.html > > The standard book on the subject is: F.J. MacWilliams and N.J.A. > Sloane, The Theory of Error-Correcting Codes > > Oscar. > -- > boykin at pobox.com http://pobox.com/~boykin ICQ: 5118680 > Key fingerprint = 159A FA02 DF12 E72F B68F 5B2D C368 3BCA 36D7 CF28 I've been doing research on erasure codes for a couple years now, and I'm working to make an efficient, non-patented variation on the tornado codes idea. Thelema -- E-mail: thelema314 at bigfoot.com If you love something, set it free. GPG 1536g/B9C5D1F7 fpr:075A A3F7 F70B 1397 345D A67E 70AA 820B A806 F95D -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 232 bytes Desc: not available URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20011010/28f5451a/attachment.pgp>
