On Tue, Nov 27, 2001 at 09:01:20PM -0500, Gianni Johansson wrote: > On Tuesday 27 November 2001 19:41, Oskar wrote: < > > > Can we use some matching of the keytype bytes instead of the length of > > the keys to determine this? > > No problem. I will fix the code to check for 0x0302 or 0x0203.
No, that is no good either. This should be exclusive, not inclusive. You should make it skip 0000 keys, and somebody should fix the announcement code to add that to it's keys. > > It is not ok to assume that that keys will > > always have any length. Any distribution of the keys should be based > > only on the first single or half byte anyways. > > Right now I am using the most significant nibble, but if keys can vary > length, I guess I should be using the least significant nibble. You should use the first byte in the array, and the most significant bits of that. Anything else will give a completely meaningless distribution. (Don't ask me which is considered most or least significant, or anything about the order of big-endian or small-endian etc - something about my brain renders me unable to remember binary spacial distinctions. Don't even bother telling me which is which, or which is used or whatever, because I won't remember.) > > Gianni: How are you making the Histograms? > Just printling ascii chars. If that's not pretty enough, people can > grab the raw data with the /keylist.txt URL and use the plotting program > of their choice. You should at least give "gnuplot" type output then. > >I know of one set of > > Histograming tools under the FreeHep program, called Aida - > > http://java.freehep.org/ - though I doubt it works under 1.1 as graphics > > was generally bad before 1.2... > I am not worried about making it look pretty at this point. The important > thing is to get some data and figure out what's going on with the network. The network is degenerate. You will see no clear distributions of keys anywhere. You really didn't need to go to all this trouble to figure that out. <> -- Oskar Sandberg oskar at freenetproject.org _______________________________________________ Devl mailing list Devl at freenetproject.org http://lists.freenetproject.org/mailman/listinfo/devl
