His argument is that the "new paradigm" is constantly changing and reinventing itself (usually with old concepts). Meaning in ~2 years, there is going to be a "better" way to do things. I'm not sure I agree with him, but that's his argument for not switching to a front-end MV* framework.
He seems to be in the "if it works now, why change it?" camp. So at the moment that means using Java Server Faces on the front end. I haven't had much exposure to the server-side, but I'm so not sure there is any good framework in place as it's closely coupled with the front end logic.
Your second comment is a good one. What I think he meant by "trend" was that last year Rails was what all the architects and leaders at the company were pushing for. His argument was that this year it's the same thing except with single page architecture (Backbonejs in this case).
From an entirely outside experience, I saw at the company I work for a very bad experience with a project where the team pushed for a single page application - they got delayed for months because of the technology, the problem was not single page applications in themselves, but that they seriously underestimated the learning curve and they were absolute novices (no-one had prior experience in SPA). And I definitely believe they could have cut months from the development time by sticking to known tech.
For a company where development is not core, I wouldn't recommend it, but if the company is a tech company, I would evaluate devoting resources to learning single page applications and doing some pilot projects, then leveraging the experience and pushing it into the main projects if applicable :) .
Very cool. Scanning the source code makes it look like you were using Alpha-Beta pruning[0], but I'm not totally sure why the function is named 'pvs'. Does anyone have a quick explanation?
