>It might be nice if the FEC download, after, say, 10 minutes of 
>requesting, looks at the failure-to-success rate and estimated whether 
>enough blocks are likely to be retrieved to reconstruct the file.
>
>This could save people a-lot of wasted time downloading blocks for a 
>file that they are unlikely to be able to reconstruct.

personally i don't think ihis is a good idea, as it will vote a successful 
download on it's transfer rate, rather on possibly available blocks.
if you have a very scattered file, which has been around for some time, it will 
require some time to find the chunks, but (assume) eventually they are all 
reachable. thus if you decide on download speed if the download could be 
successful, you effectively make all files unretrievable if they can only be 
found 
with a time consuming htl of 25

also, once these files are detected as not-retrievable, a) the other chunks are 
not dragged though freenet, b) no healing wil occur, c) no chance of a 
successful download unless *someone else* has more luck retrieving the 
splitfile, which will then increase thew speed of *your* blocks, which will 
enable the 
download for *you*, without having any control about the hardness to search for 
the chunks.

pro:
- the user's minimized splitfile window will spit out a "failed" result faster

cons:
- a), b) and c) all reduce splitfile availability
- download of a splitfile is aborted, although the file could have been 
retrieved

ergo:
- either so not estimate on download speed but on success/DNF ratio (but then 
again, this is still stupid: why add checkblocks when be cancel the 
download prematurely?)
- incorporate that *COOL* transfer page (like opera has one) so the user does 
not have a new window for every download, so the number of open windows 
is reduced
- perhaps give a %-success estimation for user's information and the download 
success chances, but DO NOT cancel the download automatically based on 
the value!

comments?






_______________________________________________
devl mailing list
devl at freenetproject.org
http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl

Reply via email to