About reinventing the wheel
Nice article: I Can't Wait for NoSQL to Die. Fits on many IT problems, not only SQL vs. NoSQL. Should be common knowledge, but apparently is not.
Bottom line: First of all do some research before throwing out the baby with the bath. And no: Jumping on the current hype won't magically cure your self made skill/knowledge/management/whatever problems.
Comments
Post a Comment
This site's webmaster failed misserably in upgrading the underlaying web framework.
The comment system is entirely broken right now. Sorry!
I'm not sure your interpretation of the article is right, actually. I read it mostly as a rant about NoSQL being an outright bad idea, whereas you're drawing a more general advice about not rushing to adopt tools you don't need. Something I'd entirely agree with.
For what it's worth, I think NoSQL systems have their place - but it's not a very big place. The number of organisations that need that kind of scalability but can't get it from a traditional relational DB - well, let's just say it's very small.
One observation, too - several of those NoSQL system are actually built on top of traditional SQL engines, often MySQL, where they benefit from a well-tuned indexing system, even if they're using rather exotic data models on top of it...
Your "Bottom Line" is nice that I want to use as an email-signature.
Simon: NoSQL obviously was the author's motivation for writing that rant, still that linked article containts quite some general wisdom. Just replace the words "NoSQL" and "SQL" with your favorite competing technologies and see your self.
Well, and I am not judging a bit about NoSQL, although I'd like to mention that key-value stores are nothing new. Would have to do some research, but key-value stores are that obvious that they should actually predate SQL databases. For instance Berkeley DB is a very mature and robust implementation of this NoSQL concept that's in production use for decades now.
Well, but also I wonder if it's really useful to entirely give up schemas. For instance RDF tripple stores also are a quite impressive technique for storing data, when SQL becomes limiting.
To keep generalizing, a lot of the "NoNoSQL" arguments sound to me like the arguments made against Linux a dozen years back. How proven commercial Unices were the thing and how the hype wouldn't solve anything...
I think it's just very, very good to have choice and excitement in this "layer".
Maybe "NoSQL" can be associated with strange, simple-sounding promises but it gets people excited, just like Linux did, and that's actually a pretty fine way to eventually solve the hard problems.
The name though is a real bummer, as it kind of forces "SQL people" to react...
Stefan: I don't think you can compare Linux with NoSQL.