>From Matthew Toseland <[EMAIL PROTECTED]>
>> Clients have to insert keyword indices for keyword searches to work
>> (doh!). The two methods mentioned in David Wentzlaff's (w/Likuo Lin &
>> Alexander Yip) searching paper at
>>
>> http://cag.lcs.mit.edu/~wentzlaf/classes/6.899/project/public/doc/
>>
>This is an interesting paper. Are LIFs evil or just hard? They seem to be
>similar to updating, would they be implemented together?
The above document relies heavily on keyspace "probes" (determining if a file
exists without downloading it) and KSK enumeration (naming keys according to a
pattern like KSK@name#0, KSK@name#1, ...). I'm under the impression that both
of these were unacceptable for security reasons.
If we had secure node-side support LIFs (instead of enumeration), and they
contained digitial signatures to allow for spam filtering, then the only
difficulty would be that filtered out "spam" replies would continue to
propagate and take up bandwidth. This could be fixed by requesting keys by
(keyword, digital-signature-list) pairs, but the digital-signature-list part
is likely to be far too large to be an acceptable search request (160-bit key
* 10,000 nodes in trust tree = 200KB upload per search).
I'm not sure if I'm hopelessly mistaken or making any sense, but I think
that's the current problem with implementing searches.
--
Benjamin Coates
_______________________________________________
Freenet-dev mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/mailman/listinfo/freenet-dev
