On Wed, Aug 6, 2008 at 11:50 PM, Michael Rogers <[EMAIL PROTECTED]> wrote:
> Ed Tomlinson wrote:
>>> Do we need to store the location for each key? (so that we can update
>>> the average location when the key is removed)
>>> Maybe a running average is just good enough?
>>> Or just store an approximated location, not the actual value?
>>
>> A salted location would work as long as the salt had a mean of 0 and was not
>> all that large (eg. less that the routingMissDistance value).
>
> I believe the salt is added before hashing, so the salted-and-hashed key
> is effectively random WRT to original key.
>
>> I this case, I think we _will_ get more meanfull results from a true average.
>
> I agree. Storing the location as a double (which is what we're going to
> convert it into anyway, so no point storing the full key) only requires
> an extra 8 bytes per record, right?

We digest the key in data store to prevent enumeration of keys in the store.
If we store the actual location, it would defect this purpose..
I suggest storing the location as float, which is only 4 bytes and may
enumeration much more difficult.

> Cheers,
> Michael
> _______________________________________________
> Devl mailing list
> Devl@freenetproject.org
> http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
>
_______________________________________________
Devl mailing list
Devl@freenetproject.org
http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to