On Fri, Mar 21, 2003 at 02:00:32AM -0800, [EMAIL PROTECTED] wrote: > > On Thu, 17 Mar 2003 20:09 -0800 Ian Clarke <[EMAIL PROTECTED]> wrote: > > >> I had another idea. How about to do some weighting towards smaller > >> keys? > > > > We actually had something like this in 0.3, but it was removed in 0.4 > > (for reasons that I still don't understand). > > > > It worked by inserting larger content further down in the LRU. For > > > example, if a 1MB document was added to the DS, then it would be > > inserted in the datastore low down enough such that there is at least > > 1MB of content above it. > > I vote that the nodes can do that weighting towards smaller keys. > > My idea is to calculate the modified LRU: > > LRU_mod(key) := m(keySize) * LRU(key). > > Where m(.) is defined by a function or by a table like for example: > > keySize | m(keySize) > --------+----------- > 1k | 1.000 > 2k | 0.999 > 4k | 0.997 > : | : > : | : > 256M | 0.850 > 512M | 0.840 > 1G+| 0.820 > > Advantage: I (as a node owner) can place a little emphasis on "content" > (html, txt, images, freesites, ...) and delete "warez" ;) (isos, DVD- > rips, music) a little bit sooner than usual.
Inserted Warez would simply use smaller splitfile block sizes. > > --Palomitas de Maíz -- Matthew Toseland [EMAIL PROTECTED]/[EMAIL PROTECTED] Full time freenet hacker. http://freenetproject.org/ Freenet Distribution Node (temporary) at http://80-192-4-36.cable.ubr09.na.blueyonder.co.uk:8889/eOlipDdDWXs/ ICTHUS.
pgp00000.pgp
Description: PGP signature