A couple of wiki-related software-development notions crossed my consciousness in recent days. Lest I lose them:

  • Google-page-rank experiments: Rumor hath it that the Google search engine uses a relaxation-method algorithm to identify "better" or "more authoritative" web pages. In brief, take a network of mutually interlinked pages; give each one a weight of 1.0; then repeatedly let each page keep a fraction 15% of its former weight and distribute the remaining 85% equally among all the pages that this page has links to. (The fraction 15% is arbitrary, but something in that ballpark is reasonable.) Then repeat the process until all the weights converge to a stable set of values. If this procedure were applied internally to the ZhurnalWiki, what would be the result? My guess is that the pages ZhurnalWiki and perhaps FindPage would get weights in the ~100 area, since everybody else links to them via standard Wiki headers and footers. A few "popular" pages like MarkZimmermann and certain "Topic" pages (e.g., TopicPersonalHistory, TopicScience, TopicSociety, ...) would also be heavily weighted. Ordinary wiki pages would probably settle down to a weight of ~0.5 or so. Some among them might do better (perhaps My Religion and My Business, two of my personal favorites). I guestimate that with a few hundred lines of Perl I could do this calculation, a similar effort to the CorrelOracle's automagical linkages among ZhurnalWiki pages.
  • Graphical-diagram-generation experiments: Wouldn't it be spiffy to generate a pictorial representation of the ZhurnalWiki, with nodes for each wiki page and lines showing the links among pages? This was a deep-back-burner concept until recently, when I was typing back-and-forth with a family friend (FH) student at UMBC. I mentioned the fantasy to him, and he pointed me to — which in turn led me to where I began to read about the wonders of Graphviz and the associated Dot and Webdot software. Alas, after an hour or two of messing around I haven't yet gotten these programs to actually run properly on my computer (I'm having problems with UNIX paths, I fear).

Perhaps I can make some progress on these in a few months, if there's a huge blizzard this winter that keeps me home for a week (but doesn't cut off electrical and telephone service) ...

(see also PetBibli1 (23 May 2000) re an article I wrote ca. 1979 on relaxation methods to solve Laplace and Poisson equations; see IrWishes (4 Jan 2001) for notes on information retrieval programming project plans, mostly unrealized still with the exception of the autolinker; see for a snapshot tabulation of the number of links pointing into each ZhurnalWiki page, not including referrals via wiki headers and footers)

TopicProgramming - TopicZhurnal - 2003-11-14

(correlates: ChatTuringTest, Periodic Tables, CloserToFine, ...)