|
|
|
|
|
|
by snorkel
6195 days ago
|
|
|
I disagree with slides 20 - 30 (calling "entity-attribute-value" tables an antipattern). Variable attributes table not only gives you a lot of application flexibility but also helps performance in that you're not polluting your primary table with extraneous metadata columns that are hardly ever referenced. It doesn't surprise me that DB programmer would look at that and be offended that it's not purely referential -- not everything that happens in a database has to be referential to everything else. Sometimes databases are just dumb data stores. Not all problems are solved by "referential integrity" |
|
|
RDF would have been great if the W3C wasn't full of idiot blowhards.
There are some application domains that have not lived up to their potential because of RDBMS abuse where EAV would fit perfectly -- Desktop Search for example. All of [Spotlight, Google Desktop, Beagle, Gnome Tracker] use RDBMSs with a pre-determined set of 'attributes' baked in, crippling any creative use. Spotlight at least has the decency to let you add your own non-indexed properties, but it's not very useful. The contents of people's filesystems and metadata about them will never fit into some spec-writer's schema.
Compared to RDBMSs I think sugared EAV is lot more amenable to being abused in the "when all you have is a hammer" way -- they start a bit slow but scale linearly with more resources without getting any slower.