not falling down

It took the Ben Folds cover of “Bitches Ain’t Shit” to make me understand that it’s a ballad about the breaking of a friendship, love betrayed and the loss of innocence. Also some pretty bizarre pronoun mappings. Consider:

“I used to know a bitch named Eric Wright

We used to roll around and fuck the hoes at night

Tighter than a muthafucka with the gangsta beats

And we was ballin’ on the muthafuckin’ Compton streets

Peep, the shit got deep and it was on

Numba 1 song after numba 1 song

Long as my muthafuckin’ pockets was fat

I didn’t give a fuck where the bitch was at

But she was hangin’ with a white bitch doin’ the shit she do

Suckin’ on his dick just to get a buck or 2

And the few ends she got didn’t mean nothin’

Now she’s suing cuz the shit she be doin’ ain’t shit

Bitch can’t hang with the streets, she found herself short

So now she’s takin’ me to court”

This all refers, of course, to the feud between Eazy E and Dr. Dre that was associated with the breakup of NWA (longtime readers may remember my fascination with Eazy’s response to the album). To parse it correctly you have to understand that except for the passingly-mentioned “hoes,” everyone referenced is male.

The gender issues involved in early 90s gangsta are a Master’s thesis all their own. (Actually, I bet they already have been, cheap Berkeley joke here.)

I just had a stupiphany that would have been great if it had happened six months ago, when I had the option of submitting a thesis. Whoops!

See, you could store bitmap indices on a varchar field in a database as a two-dimensional black-and-white image for each character position! Normally you wouldn’t use bitmaps for text because the density would be less than 1%, but what if you compressed them like PNGs? You could save a huge amount of disk space, because sparsity would improve the compression ratio, and clock cycles aren’t nearly as valuable as disk access in this kind of situation so the decompression would be parallelizable. The binary operations would reduce wildcard search time by like an order of magnitude! Hell, you could probably store an average-value flattened composite for returning more relevant results faster, and since varchars only go to 256 you could do it as a standard grayscale image! And imagine the data mining you could do on a map like that, pattern recognition, domain linguistics, not to mention just rasterizing it and putting it on your wall…

I promise there are people out there to whom this makes sense. Oh well. Maybe someday I’ll write a paper.

The motherfucking Compton streets.

I’ve gotten bored with Dog Bites, so I’m probably gonna quit posting there for a while unless something blah blah blah. This is what RSS feeds are for!