When Your Customer is Your Competitor: The Return of Roll Your Own

True, but then there are situations like Rails, where 37 Signals could have built their apps using existing tech, but said 'there should be a better way'. I mean, if you told someone "I'm building this project management app, and I'm going to take an esoteric language no one knows, and write an app framework from scratch" they'd rightly tell you you're crazy.

Most endeavours that start like that end up as failures, however without these attempts tech would stagnate.

Disclosure: I wrote the piece.

But anyway, I'm still with andrewvc here. I concur that NIH - the pathology you're describing above - is equally unproductive, and that deciding when to build or buy is a difficult question to answer.

That all being said, it seems fairly clear that due to a wide variety of factors, heterogeneity is accelerating, which makes life more difficult for general purpose, volume solutions. Think of it this way: Windows XP was, for a lot of people, a more than acceptable compute interface. But it wasn't designed for phones, devices, tablets, and the like. And attempts to bend it to that purpose largely failed.

We need different tools for different jobs, and with the number of job increasing, sometimes there just isn't a good tool available. At which point rolling your own is not only appropriate, but desirable.

Point being that specialization is the new norm, and as such roll your own - or at least leveraging the results of someone else's roll your own - will become more common. IMHO, and so on.