June 16th, 2002

Trippy, tired, Tired


Spent most the day working on the path with Nick ... excavated 4" of dirt 3.5 feet wide around the house, put down anti-weed paper, covered it in sand, and started bricking. Lot of work.

Nothing really exciting going on ... I have some things I want to work on, but need to go dig out a book from storage before I proceed.

Moving in about a week... between the 22nd and 24th? Who knows. Need to figure out what's going on tomorrow when people are around. I guess the bank did their appraisal... just drove by and said 'yup'. So all's a go ... just need to start signing things like crazy.

My email is depressing ... why can't people just be cool?


fun problem

Something I've been musing over....

Given a predicate calculus sentence, along with the number of times that each predicate literal and its negative appear in the search space, rearrange the sentence such that it's logically equivalent, but more likely faster to evaluate for truth.


p OR q

If 'p' occurs 2% of the time, and 'q' occurs 90% of the time, it'd be quicker to test "q OR p" over 1,000 cases, instead of using the original. (we're assuming the cost here is looking up the literal values, not doing the AND/OR)

So, I guess it's not too hard....

In a series of disjunctions, move the highest probability disjunct to the beginning.

In a series of conjunctions, move the lowest probability conjunct to the beginning.

For clauses more complex than a literal or negative literal, I could fake the probability ... for a conjunction, just multiple the probabilities. In a closed search space, that's nowhere close to the truth, but it'd make a good heuristic. After all, it doesn't matter what weightings we use, really, as long as all transformations keep the logical equivalence of the original sentence.

I'm probably describing this entirely wrong. I want to go dig into storage and get out some of my intro CS books so I could at least describe things using the right terminology.

(no) fun with NFS

Today's involved fucking around with NFS, way more than I wanted to.

Let me summarize my hours of research:

-- NFSv2 is dumb.
-- NFSv3 is dumb with large file support.
-- WebNFS is an attempt to make NFSv3 not dumb.
-- NFSv4 isn't so dumb, but is still in development.

Why is NFSv3 dumb?

-- 3 different services, many different ports
-- TCP and/or UDP in places
-- can't load-balance/firewall easily
-- encodes host name of client in mountd request (!?)

Still fighting it all.

Might cave and go with experimental NFSv4 support. *gasp*

Time for food.