On Tue, 16 Apr 2002, Will Glynn wrote:

> > That equation only holds if the probability of failure for each block
> > are independent from one another which is certainly not the case.
> 
> Could you explain, exactly, why failure of a piece of data is not
> independent from failure of another block? Unless nodes keep tabs on
> individual blocks of data (as referenced from index blocks, or
> whatever), failure of a single chunk of data is in no way related to
> failure of another chunk of data.

For one since the blocks are part of the same file if one block is 
downloaded there is a good chance the other blocks will be downloaded 
also.  This means if one block is highly popular so will the other 
blocks.  Which means the availability of both blocks will be the same so 
once you have successfully download a few blocks without failure this means 
the data is popular enough to be readably available, so the other blocks 
are extremely likely to also be available.  I can't prove this will be the 
case, this is just what my intuition is telling me.  If that is not 
enough I can add extra consistency checks just as making sure that the 
network doesn't discard random blocks of a file without discarding them 
all.

Also, data retention is going to be a *lot* better on by network so the 
chance of failure is also going to a be a lot lower than on freenet.

--- 
http://kevin.atkinson.dhs.org


_______________________________________________
freenet-tech mailing list
[EMAIL PROTECTED]
http://lists.freenetproject.org/mailman/listinfo/tech

Reply via email to