Oskar Sandberg writes:
On Thu, 13 Apr 2000, hal at finney.org wrote:
The important thing about comparing strings of different lengths
is the difference in length should have as low signifigance
as possible... Other then that it doesn`t really matter, the
non-intuitivity of the current method stems mostly from the
non-intuitive layout of ASCII...
Actually, what I find most non-intuitive is that the search for the next
node essentially moves in "both directions" (both lexicographically
up and down) simultaneously. Aside from the issue of whether "ab" is
closer to "aa" or "ac", this makes it hard to simply enumerate the keys
that need to be searched. If the next node moves in one direction, key
search can simply be expressed as sorting the list of keys, finding the
mask, and getting the next key. Using a sorted list reduces the time
required to find the keys from n squared to n log n.
I guess the same can be done with the current closeness metric, it's
just that the sort function is so much stranger...
_______________________________________________
Freenet-dev mailing list
Freenet-dev at lists.sourceforge.net
http://lists.sourceforge.net/mailman/listinfo/freenet-dev
