RailGunner's Journal: Beta, Part 2

You know... since I've written enterprise applications (singlehandedly, and no, I won't tell you which ones) that dwarf slashdot... I might be tempted to write a replacement, given how bad Beta is and how much people are complaining about it...

How would I do it? In the spirit of Open Source, I'd use PostgreSQL as a database backend, as MySQL or MariaDB is, in my humble opinion, inferior to PostgreSQL.

One reason is the syntax of PL-PGSQL vs. T-SQL. I hate T-SQL compared to PL-PGSQL (or just PL-SQL if I'm writing a Stored Proc for Oracle).

After that, I'd model the database to do what I want, and store what I want, and I'd ensure that the data was fully in 3rd Normal Form.

And if you bitch about joins, I'll hit you with a fucking sledgehammer.

After that, throwing up some PHP to format the page is an afterthought. The critical piece is getting the database right.

The database itself would need to be clustered to be able to scale with the load, and PostgreSQL does a pretty nice job of it.

3rd Normal

[smittyoneeach]

But have you gone after Big Data, though, where supposedly all that indexing triggers butthurt? Asking for a Hadoop friend.

Re:

[RailGunner]

Yes, as a matter of fact, I have. BigData is a useful tool, and one I've used to varying degrees of success (used MongoDB, Hive, HBase, CouchBase (ugh)). If you want to discuss it, I'll talk more freely in email. naqamel at gmail -- I check that account fairly regularly.

And yes, for something like Hive -- you don't want to normalize data. Duplication is OK, as it's operating under the assumption that disk space is cheap, and you're interested more in raw speed than in referential integrity. Not the best f