typical stuff to do in dbi:
	connect
	insert
	select
	delete

a common method for abstraction is to end up with wrappers around DBI, and this
can end up making your life more complicated than the SQL did to begin with

models
	what are they
	why do db's suck at it?

a model is a collection of things
like a recliner:
	covering
	height
	weight
	position
	age

databases suck at this kind of thing
	they have tables and columns and records
	they make us manipulate recordsets, but we really want to manipulate stuff
	(models)

we see things as objects
	with properties
	and relationships
	and constraints
	and "mortal traits" [ed: they can die, I think]

objects
	answer questions
	and do what we say

	object-relational mappers (ORM)
	wrap your existing relational schemas
	remove SQL from applications
	(remove SQL from the code you'd otherwise enjoy writing)

we need to configure:
	db connections
		classes rarely share
		but they do share these
	table spec
		you can spec what sequence (a la pgsql) to use
	column categorization
		five types:
			all
			primary   - the key
			essential - always fetched
			others    - retrieved lazily
			TEMP - non-persistent
	relationships
		has one
		has many
		might have - inherit behavior (a la pgsql inh. tables)

better searching than builtin
	AbstractSearch (see SQL::AS)
	custom constructors
		performs search, returns set
		you can use placeholders to parameterize [ed: very cool!]

class::dbi is just a platform on which to build
you don't want to just use it with no optimization