Saturday, November 17, 2007

Beware the Hail Mary Meta-Hyper-Framework

A while back I spoke with the CEO of a midsize enterprise software company. He invited me to work with a small team building a framework for customers to create applications that would run with his company's app.

I spent some time with the team, looked at the framework, and declined. As diplomatically as possible, I explained what I saw as the key problems with the initiative:

  • The term "architecture astronauts" barely touched these guys. I think they actually used the term 6GL in conversation. They were architecture hyperspace/time-travelers.
  • At the same time, I'd call them "infrastructure spelunkers" as well. The team trusted nothing, and was reinventing everything. They were rebuilding so far down the stack I started wondering where they built their own hardware. Then I realized that was silly -- after all, they would mine their own copper before they'd use someone else's wires.
  • The initiative had been running for several years with no concrete delivery. There was no deadline for a future delivery either. Given that the predominant tools and platforms were evolving underneath this team faster than the team was moving, I suggested I did not expect they would ever release.
  • No real world app -- either inside the company or among its customers -- was using it yet. That is, there was no dogfooding or validation of any kind that the framework was learnable or made the relevant problems easier to solve than existing approaches.

The CEO appreciated my frankness, and said that from his point of view, it was a big long-shot bet:

  • For him, the 15+ man-years that it cost to work on this framework were a non-issue. With plenty of revenue and people, if this initiative went nowhere, it would not be a significant loss to the company.
  • The framework was not critical-path for anything else, so he didn't anticipate any major downside.
  • Potential upside was, he felt, huge. If this framework somehow succeeded, he could grow his company much faster than he could any other way he could envision.

Well, I thought, I wouldn't make the same choice -- I would probably look to gain the same leverage using a radically simpler framework or mechanism -- but then I wasn't looking to take on an executive role, so it wasn't really my call.

DHH, the man behind the Rails web app framework, wrote about frameworks when he explained "Why there's no Rails Inc":

"... Rails is not my job. I don't want it to be my job. The best frameworks are in my opinion extracted, not envisioned. And the best way to extract is first to actually do. ... That's really hard if your full-time job is just the extraction part since you now have to come up with contrived examples or merely live off the short bursts of consulting. For some that might work, but I find that all my best ideas and APIs come from working on a real project for a sustained period of time."

Nicely put.

No comments: