[RC5] Date: Mon, 11 Feb 2002 21:10:38 -0000

daniel carter hedonist at win.co.nz
Wed Feb 13 00:49:10 EST 2002


EVIL|Raven|RoT wrote:

>hmm I don't want to sound pessimistic but actually it could be possible. Let
>say we have a massively overclocked cpu that finishes a block right in the
>moment when it crashes due to overheating (or a little earlier that the
>block gets into a proxy) then it could be that the right key is marked as
>wrong.
>
If it crashes as it finishes a block.  Then either:

1) it crashes before writing the block to the out buffer
2) it crashes as it's writing the block
3) it crashes after writing

1)The block will be continued if checkpointing is enabled from the last 
checkpoint, or simply lost if checkpointing is not turned on.
2)The out buffer will be corrupted due to a partial write, and the block 
will not be sent back as it did not get fully written to the out buffer.
3)Assuming the change to the file is flushed from the OSes cache then 
the block will returned correctly.  If the write was not flushed from 
the OSes cache then see 1.

>
>The possibility is quite small but there is one.
>
>Or dnetc could make a mistake during the recycle process (hmm also a quite
>small possibuility)
>
Or a massive disaster in america could destory d.nets servers and all 
backups...


--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list