[metaforth]

Neat.

One of the arguments I hear against Python is that while initially things happen fast, the effect does not scale for large applications. Or that Java/C#'s static type checking makes it more scalable.

I'm curious if you have seen any scalability effects. For example, did the Python teams productivity stay consistent or did it taper off the further into the project you got? Did the C# team's productivity go up over time? Stay constant? Drop?

I'd be really curious to see the time series on the productivity.

[bjelkeman-again]

Yes, that would be interesting. I wonder how much this depends on how much "low level" work one has to do though? In the article he talked about the programmers not being that familiar with Django yet. It depends of course a lot on what you are building, but in our development team we find that we don't touch Python that often and mostly build at the Django layer.

[kgrandis]

Both really good questions. That was started almost two years ago so we've worked on a variety of different types of projects since that initial study--from simple websites to high-availability, rich applications. Some have required more internal knowledge of Django than others, but I can't think of a time where Django itself was an impediment or a bottleneck.

With regards to time series...me too! I don't have enough data to make any statistically-backed claims, but anecdotally Django-based sprints do start off with a big bump in velocity, because you're handed a tremendous amount of functionality upfront via the framework itself, Admin CRUD, etc.. Once a team gets moving their velocities tend to stabilize, which is one of the reasons story points and velocities are such powerful estimation and forecasting tools.

That being said, C# teams seem to have larger standard deviations around their mean velocity. One thing I have noticed is that refactors tend to have differing impacts. Major refactor efforts and rearchitecting almost always noticeably hit the C# team's velocity much more so than the impact on the Python group. Having done refactoring and and rearchitecting in both environments, I tend to think it's a function of the syntactical overhead associated with the C# language.