Sunday, September 21, 2014

Terry Pratchett, the Long Earth and IT projects

I’ve just finished reading the “The Long Earth” novel by Terry Pratchett and Steven Baxter. While I expected the book to be more Pratchett’ish in the sense of humour (it’s not) I found it a very inspiring read anyway.

Parallel worlds

“Long Earth” is based on the idea of parallel worlds. While reading it, I couldn’t resist the temptation of imagining what would happen if we had the same idea applied to IT projects.

Remember the last time you started a project? I’m sure you’ve had some ideas on which architecture to choose or more likely, which language/framework to base it on.

Should we go with Rails? Isn’t node.js the better thing nowadays? Is it a good idea to go with CQRS here? Scala will save us? This time it will be better?

Now, imagine that all the decisions were made and all the alternative realities are happening.

Let’s say, that in the Primary World you went with Rails, Angular, without CQRS.

3 months in, you have the typical Rails problems, the models are overloaded with responsibilities, the controllers are not really that thin. The build now takes 5 minutes while most of the frontend code is untested. The team members start complaining - “We wouldn’t have such problems with Node.js”, “Can you see it now - we should have gone with CQRS in the first place”, “Scala, we need Scala, because of reasons (monads)”.

Given the idea of parallel worlds (or rather parallel IT projects), you could jump into the world where Scala was chosen, just to notice, that the project is even in a worse shape, because some people didn’t get the idea of non-mutating state and it’s all not pure enough, while the frontend is not done at all, because all the time was spent discussing why Scala is actually worse than Haskell.

Then you jump into the node.js alternative world. A quick look at the codebase - yes, callbacks everywhere. How did it go with code reuse on server/frontend? Oh yes, they don’t really share that much. Also, part of the codebase is in JS, while the other parts are in CoffeeScript, because you hired those new JavaScript developers, who hate Coffee, while the Rubyists in the team can’t parse JS at all.

Wouldn’t that be great?

It’s all science-fiction obviously. I wonder, though, how much would it change our decision making process, if we knew that we could always look at how it goes in the alternative worlds.

Would it make our decisions better?

Would we learn that all projects have problems at some point, despite the technologies and architecture choices?

Would we learn that no matter what we do, we may end up with another legacy codebase?

Now, back to reality. We have only one, the Primary World. We can’t jump into other worlds. What tools do we have to measure our decisions? Is it actually worth to measure at such a high level?

Maybe it’s better to just learn how to get out of any mess that we (and others) create?