Monday, November 29, 2010

Black Friday Buys

Picked up a wireless access point for 10 bucks on Placed it at the highest point in the house, and set it up to pull it's own IP address from DHCP off my existing wireless access point, so it's basically a range extender. For a $10 "TRENDnet", I was really impressed at how easy it was to setup. I tried the same thing with a Linksys several months back. The linksys would work on and off and just periodically die, and it cost something like $50. Anyway, I'm posting from my living room and receiving a good signal, so definitely worth the $10.

Saturday, November 27, 2010

NoSQL: Quite Possibly The Stupidest Catch Phrase Ever?

A recent article on Gigaom is titled "Why cloud computing sells and NoSQL is fading".  I'd like to parlay that discussion into one of "why NoSQL is quite possibly the stupidest catch phrase ever". Here's where I'm coming from. Clearly the "NoSQL" moniker arose from a desire to describe systems that avoided the scalability limits of relational databases. But from the get-go, the term is just...STUPID. First of all, SQL is a query language, and there are relational systems that don't use SQL. So, "NoSQL" isn't even a good way to say "non-relational". Secondly, defining systems in terms of what they *don't* do means that dozens of systems that shouldn't get grouped together, do get grouped together. Corn: NoSQL! Windbreakers: NoSQL!

Adding the confusion was this classic: Map Reduce: A Major Step Backwards.  It's amazing that this was written by Michael Stonebreaker, but this just goes to show you that even a genius can write a ridiculous article when he's writing for a PR blog. Then there was the 180 degree follow up (ooops, how stupid do we look?): Reaffirming Our Commitment to Hadoop and Mapreduce.

Honestly, the entire NoSQL thing is, to me, a mark of short sited PR as well as a failure to recognize that SQL databases can and must peacefully co-exist with systems designed to tackle other aspects of scalability and data processing. When Hadoop and Mapreduce get thrown in with "NoSQL" it especially drives me nuts, because Hadoop and Mapreduce are not even data retrieval or management systems! C'mon folks, Hadoop is for data processing. And there are lots of reasons to take the output and put it in a relational database. 

So, I was happy to see the Gigaom article. The NoSQL emperor has no clothes. It's a stupid catch phrase, and a clear indicator that the person using it is clueless.