[BitTorrent] Back to Merkle Hash Trees...

Olaf van der Spek OvdSpek at LIACS.NL
Sun Feb 6 19:31:54 EST 2005


Elliott Mitchell wrote:
>>From: Olaf van der Spek <OvdSpek at LIACS.NL>
>>Elliott Mitchell wrote:
>>
>>>>From: Olaf van der Spek <OvdSpek at LIACS.NL>
>>>>Elliott Mitchell wrote:
> 
> 
>>>one hash at a time you've generated thousands of extra protocol messages.
>>>Given this you're likely to transfer blocks of hashes, at which point
>>>blocks of hashes will likely be of similar size to payload blocks.
>>
>>But with a different block/field/message type.
> 
> 
> My ideas have been otherwise, simply handling blocks of hashes as any
> other block of data. Pointing to payload versus internal nodes implicitly
> via their indicies, but otherwise identical to any other data. My
> thoughts have also not been constrained by thinking of "THEX", so I may
> of been exploring cases that didn't occur in your scenarios.

Maybe.

> Having said that I think my ideas have solidified sufficiently that I may
> write them up in both code and spec form soon. At which point it will be
> time to see whose ideas get shot down.
> 
> 
>>>>It should be clear from the protocol.
>>>
>>>
>>>Which is another reason I dislike THEX. The type is a protocol issue, not
>>>a standards issue. You might settle on markers similar THEX, at which
>>>point you've got to try both markers to distinguish the types.
>>
>>Why can't you add both the markers and something else?
> 
> 
> Inefficiency. If one fails is there any reason to believe the other will
> survive? In Cryptography adding a constant payload is a big no-no, as it

Doesn't that apply only to encryption and not to secure hashing?

> aids cryptanalysis. As lower levels (TCP) will take care of errors, why
> add redundancy at the application layer?

Maybe because the TCP checksum isn't (very) strong.

>>>>If not, an internal node is 21 or 41 bytes (IIRC), so your 16 kbyte 
>>>>block would not be an internal node.
>>>
>>>
>>>And a payload block can't be of this size? (notably the EOF)
>>>
>>>You're still assuming transfer of one hash at a time, which is worthless
>>
>>Am I?
> 
> 
> Time for those write-ups.

I doubt those are gonna say anything about my assumptions.


 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/BitTorrent/

<*> To unsubscribe from this group, send an email to:
    BitTorrent-unsubscribe at yahoogroups.com

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/
 





More information about the BitTorrent mailing list