And some Social software tends to lend itself to personalization through recommendations, so Apache Mahout (more precisely Taste) make sense there.

We have been waist-deep into those problems for over a year now. And, ironically, after Economist’s article last week I made the opposite conclusion – we have to give up the idea of re-using & integrating of various open-source libraries and do it on our own as proprietary software.

We have whirled significant amount of open-source projects over the year (including many of the mentioned in this article). The reasons we are partially giving up with integration effort are simple:

1.Most of the open-source solutions are not lean – they try to solve ‘a bit more’ then necessary. As a result, there is significant functions overlap in all pieces of the puzzle. And so you just really need a good riffling machine to make it fit.

2.Good deal of open-source projects is built with ‘proprietary’ way of thinking – authors assume that the solutions would be used as-is, without the need to integrate or expand. Some of the smaller projects do not follow the patterns and/or programming guidelines, which makes the tweaking a terrifying experience. I am afraid, often times it is made on purpose – as the monetization path for open-source projects.

This said, for me, the Economist’s point was more about elaborating the standards rather than particular set of open-source packages. The open-source effort needs to be streamlined by the open-standards so that all good niche packages can then work together nicely.

My experience indicates that it would mean dumping a lot of existing OS projects and creating new ones with such ‘open mentality’ already included.

I will go ahead and update your spreadsheet though

On the other note: the stack you outlined makes perfect sense. This is exactly what we are building for our needs.

Thanks.