Multiple history files

[Bill Davidsen]

Russ was talking about multiple history files and an implementation
conforming to the API. I have a thought on that, which I doubt I'll
investigate until vacation in October...

Once upon a time I stumbled on a program called "mph" for 'Make Perfect
Hash.' The characteristics of a perfect hash are that every lookup will be
satisfied in exactly one test, with no collisions. Of course this needs to
be recalculated when the keys change (may need to be).

If multiple history files were being used, it might be possible to make a
perfect hash for the old entries, since no new entries would be added. The
payback for this CPU intensive effort is that I believe there would be NO
hash table, just a description of the perfect hash function. Sounds like a
real time and disk space saver.

Any thoughts on this, or am I remembering it wrong? I was using it to find
entries on a VERY slow disk subsystem, and it worked really well.

-- 
bill davidsen <davidsen at tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.