Debian, Berkeley DB, and AGPLv3

The choice to fork bdb seems increasingly more compelling as one reads through this article. This has to be close to the ideal situation for a fork: if there is such a thing as mature software, bdb certainly qualifies. Programs that need features found only in releases 6+ must be rare.

The only serious bump I see is symbol-name collisions, linking libraries that use different releases into one program. Linker symbol versioning could probably come to the rescue, then, and be much less work than trying to get everyone onto one bdb release.