>>but i'm quite sure that aborting the unsuccessful download prematurely will not do 
>>any good, because the last blocks that are requested for the garbage bin, 
>>because the download will not succeed due too much missing blocks, will not be 
>>requested and thus are not spread though freenet, making them drop off 
>>the net, which will lead to even more missing blocks for the splitfile and the 
>>following splitfile healing process!
>
>Yes, sure, but the pupose of Freenet is to drop unpopular info
> in favour of popular one; your argument is good to survive
> unpopular info ad disadvantage popular one.
>
>I cannot understan why FEC files must be treated opposite to non FEC ones.
>
>You are trading good user experiece for a target that IMHO
> is not valuable.

you are right; reinserting the missing splitfile blocks might increase the used 
storage space in total, because already dropped blocks are available again.
but then again; why was the healing of fecs added, when not to insure, that splitfiles 
become better chances to be retrieved successfully?

the difference between fecs and "normal one-block data" is, that you have to get all 
(or 66%) of the fec blocks to archive a successful download. but on an one-block piece 
of data, you either get a rnf/dnf and retry 
until it apprears or you give up frustated. to make a splitfile survive in freenet, 
you have to cheat a litte to keep the blocks completely available. starting a fec 
download requiring 50 blocks to be retrieved for a 
success, and then experiencing, that 4 blocks are missing, will be even worse than 
healing with reinserting the missing fec blocks, because you a) got NO successful 
download and b) shove the blocks through 
freenet without any need, because the download will never be successful again because 
of missing blocks.
this results in used store space by the remaining 46 blocks, which are in fact this: 
garbage! they only takes up space:
46 blocks * 256 kb/block = 11,5 mb
11,5 mb * ~15 nodes = 172,5 mb WASTED total in freenet stores for a splitfile missing 
1024 kb (4 blocks)
(checkblocks would take up extra space, so this calculation was done with the 
situation in mind, that all checkblocks are missing, too. you can easily recalculate 
it with checkblocks if you like)

oh, and because i requested for the file (if noone does, the fec will drop out, no 
question (no healing occurs)) i am interrested in this piece of data, so you cannot 
say it is unpopular :)







_______________________________________________
devl mailing list
[EMAIL PROTECTED]
http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl

Reply via email to