Framework First, Product Later

A few years ago, someone pointed out to me that Coding is actually a design task. That when you code the answers of how to code something are never obvious.

In contrast, someone putting the bricks down to build a house has a much more defined task. The architects for real world houses were really the only ones doing design and that for 99% of them, they were good at it only when they were essentially designing a house that they had seen most of the parts already built in the past.

While design is not very easy, it is very hard (and way too abstract) to do design^2. Humans are just not very good at it. And that we should be very careful being a software designer or as the OP suggests building a framework without a concrete product built first.

Note: When I mean software designer above, I mean someone trying to spend a large percentage of their development specing out all the details that the coders (i.e. the designers of the software world) should be working on. I am not referring to Graphic Design.

"However the patent covered migration, so no matter what, migration was the solution. We wrote the framework around the patent..."

They realized early on that there was a better way to do things, but stayed the course because they had to make something that used the patent. I don't think the framework ever had a chance from the beginning.

I blogged about the "moral" of this story a few months ago - http://rtigger.com/blog/2012/07/06/architecture-last/

TL;DR - Frameworks should emerge from use cases, not the other way around.

I was working for a company a few years ago who started building a web app in PHP. We had a new PHP developer who was very confident, who had what he claimed was a fantastic PHP MVC framework he'd written in his spare time. You can guess where this is going, but because he'd just written it as a framework instead of building it around a product and then extracting it out, the majority of the time was spent shoehorning the product to fit the framework.

I'm sure there's frameworks that exist that haven't been extracted from a product that do a perfectly fine job, but I've found that when they fall into the extracted category they've already had a round of battle hardening done, which makes life somewhat easier.

This inevitable schism between features that are nice and features that you need are why I am sitting on multiple frameworks without releasing them, because I refuse to release a framework I haven't built something with. I've had to rewrite those frameworks multiple times after realizing how broken they were in a real-world scenario.

What about frameworks like MVC .Net and Spring framework?

What codebases were these extracted from?

"If you encounter a framework that wasn’t extracted from a product, run away screaming, no matter how extensible it claims to be."

You mean Ruby on Rails? :)