We became much less restrictive about patches. Patches that fixed one bug while introducing five more were now accepted to encourage developers. We decided to accept bug reports only as MS-Word documents, which cut down on the number of bug reports. We had to do this because we were getting too many bug reports because of our new liberal patch application policy.
We stripped down the web site so lynx users could read it easier, and Palm Pilot users said their browsing was much improved. We moved our web site to WindowsNT to slow down the bug reports coming from the mailing list.
At this point, we really moved into high gear. We switched to C++ so we could use those nifty // single line comments. We added Java to some low-level routines. This slowed things down, so we randomly replaced some functions with assembly language. As the supported platform list dwindled, the requirement to have Java was not as much of a problem, and we got PostgreSQL running on a Java-enabled printer, so I think the change was worth it. Having PostgreSQL in the printer helped because we could format our source code using the database before printing it. It was really hard to read without that.
Looking back, we certainly are making progress. I am just
not sure if the progress is forwards or backwards.