Re: general gnutella security question

Thanks greg and sam for your responses. The documentation you gave me and
your reponses makes me open a discussion, at least with myself, about the
purpose of the QRP. So, now, according to this, QRP was created to avoid
some traffic with slow peers, and also of course to avoid unnecessary
queries. My first question is: why storing in the hash tables the keywords?
can I store not only the keywords but also the whole file names? I know that
to have the keywords gives more space for reponses similar or related to the
original queried string. For that reason, storing only the whole file names
give more restrains. What if I store the whole file name, and respond to a
query with a queryhit containing a file stored in another peer? would that
turn into a security issue? Now, the idea explained again:
The ultrapeer could have hash tables for its leafs. Those tables store not
only keywords but also whole file names. If a query matches the keywords in
the table of the leaf X, the ultrapeer responds to a query with a queryhit
in behalf of leaf X having one of the file names stored in X. I can't say
now how to efficiently store the file names in the ultrapeer, but could that
avoid one more step? which is the leaf X sending queryhit to the querier?
would that increase too much the traffic for an ultrapeer?
Thanx, M

On 5/12/07, [email protected] <[email protected]> wrote:
>
>   Here's the original work that QRP evolved out of:
> http://aeolusres.homestead.com/files/index.html
>
> Our early writings on the issue:
> http://www.limewire.com/developer/query_routing/keyword%20routing.htm
>
> Thanks
> -greg
>
>
> Quoting Sam Berlin <[email protected] <sberlin%40gmail.com>>:
>
> > Hi Miguel,
> >
> > > The second reason is security. Am I
> > > correct that the main purpose of hashing the key words is that other
> peers
> > > don't know the exact names of the files that a peer is sharing? In
> that
> > way,
> > > the protocol can avoid untrusted peers get knowledge of what legal or
> > ilegal
> > > files the peer is sharing? and is it for the purpose of not only
> hiding
> > > ilegal stuff, but also avoiding any kind of attacks?
> >
> > I wasn't involved in the creation of QRP, but from what I've gathered,
> > this is just a common misconception. Remember that the creation of
> > Ultrapeers (and by association, query routing tables) in the first
> > place was a response to the meltdown of Gnutella due in large part to
> > too many modem-users hindering traffic flow. This can be extended
> > also to slow machines with not much memory, and other generally
> > poor-performing factors. QRP is a lightweight mechanism that takes a
> > very large piece of information and whittles it down (as bloom filters
> > do) to a manageable piece of information, in terms of storing, looking
> > up, transferring, and more. The initial design goal was to have it
> > propagate many hops, so size and manageability were likely key. In
> > reality, the bits only traveled a single hop, which lent itself to
> > even more optimizations that would have been thrown out the window if
> > the entire listings were transferred.
> >
> > So, although from the surface it seems that security and other factors
> > might have been a part of the introduction, I don't believe they
> > actually were.
> >
> > Sam
> >
>
>  
>


[Non-text portions of this message have been removed]