Stuff Michael Meeks is doing

To give a sense of the magnitude of this interdependence, the libreoffice.org domain contains 13,281 links to webpages hosted on openoffice.org domains.

• I'm encouraged by Rob's approach - he doesn't simply re-hash this old chestnut based on some ~25 links in the wiki help (now removed), when translated to a hundred languages, for some tens of versions. That is encouraging.
• I'm amused at the 'Guinness Book of Records' precision - "13,281" is curiously precice a number, without an equivalently precise date or explanation - our website evolves and grows rapidly.
• I was discouraged by my attempts to replicate this "magnitude of interdependence". My first attempt, was to use a google link search: site:libreoffice.org link:openoffice.org - which gives me 4980 hits(today), interestingly if I search for link first I get 5170, so perhaps there is some +/- 10% thing going on.
• Digging through those links, what immediately jumps out is the sheer weight of false positives - that demonstrate no inter-dependence. Apparently our home-page and donations page link to OpenOffice.org, but I can see no trace of that. Then there are myriad mentions from bugzilla (which shouldn't be crawled I think) that helpfully cross-link to the archive of many tens of thousands of historical bugs. It also appears that our mailing lists are indexed which also here and there refer to the legacy project.
• Checking vs. link:apache.org, we get 31k hits, which starts to get to a similar ball-park, but again all the same problems - false positives, hits inside documents attached in bugzilla bugs, E-mail etc.
• It is an interesting idea to extend this idea and judge inter-dependence by looking at mutual referals. Perhaps it yields interesting results for the industry. These are webhits I took

  Relationship	A links to B/khits	B links to A/khits
  google.com vs. microsoft.com	1960	38
  oracle.com vs. ibm.com	68	21
  kde.org vs. gnome.org	30	10
  apache.org vs. freedesktop.org	41	0.5
  libreoffice.org vs. apache.org	31	7

  So - is Google overwhelmingly dependent on Microsoft ? is Oracle more dependent on IBM than vv. ? personally I'm skeptical. It really feels like this should show us something interesting, but it's an extremely crude, noisy, and misleading metric, particularly when LibreOffice development is split over two sites. I suspect the hit counts also vary wildly depending on locale, and the state of mind of the hard-working mechanical turks that produce the nice-ranking / hit count metrics.
• Some conclusions
  • Such hit counts are a very blunt and misleading tool, it is unclear if they accurately reflect any true "sense" of interdependency.
  • Inasmuch that they show some interdependency it is unclear to what extent this is just a legacy problem. It would be appreciated if Apache's hosting of the hundred-thousand plus legacy bugs, or other big chunks of historical material that they did not produce and our linkage to them are not used in this way. Hosting copies of legacy information is not particularly taxing.
  • Together these confirm my feeling that if Apache OpenOffice disappeared tomorrow, it would have little-to-no impact on LibreOffice at all, perhaps the converse is true too. To me - that is a true measure of interdependence from first principles, though of course as a project - I imagine we're eager to re-unite around a single weak-copy-left code-base and to meet IBM's legitimate needs; so perhaps one disappearing could be a positive thing.