Google groups thread

Bespin has the idea of multiple backends attaching to the same client. Right now, Python is apparently “winning” while Java has fallen behind and PHP appears to have faded away.

The situation is very analogous to Shindig (http://incubator.apache.org/shindig/) which has a strong Java back-end, a PHP back-end that took a long time to get started in earnest and AFAICT from a quick glance has tapered off, and a number of non-starters.

I do wonder whether something like this is really viable - can you have a vibrant ecosystem of back-ends in different languages? Granted, there are hundreds of implementations of SMTP, FTP, and so on, in all the languages of the rainbow. Those are actual protocols based on strong standards. Likewise, Shindig had a fair crack at it because it’s based on the OpenSocial standard (but the diversity effort is probably hampered by the fact that most companies who want to be OpenSocial hosts tend to fall in the Java camp).

Bespin would need to formalise things and establish proper standards to support multiple backends. Nothing as onerous as the IETF standards, but still, they’d need to nail down all the subtle issues and edge cases in order to support multiple backends. Good for cleanliness perhaps, but at the expense of project velocity. Years ago, I asked SpringSource’s Rod Johnston about why one should use Spring over the upcoming EJB3, and one of his reasons - quite rightly - was that his code was available NOW while the committee-driven EJB standard and subsequent implementations were a long way off. In other words, Spring demonstrates that working, open, code is in many respects far more powerful than a standard. I can’t imagine Bespin will want to formally document its protocols; instead it will rely on code. But of course, if the code is a moving target, how easy will it be for different backends to keep up?

Also, how motivated will the second, third, and fourth guys be? It’s more fun being the first to see the concept become reality; less so to be building a port from one language to another. As with the previous point, though, I stand to be corrected; some people might enjoy the engineering challenge of building a parallel backend. (Maybe I’m being solipsistic here, as I’m personally 20x more motivated by user stories than implementation details.)

Java hosting is notoriously painful compared to the ease of cheap PHP hosting, so I can see the argument for both; companies will want to host Bespin on their own shiny enterprise Java servers, while rebellious little startups will want to do the same thing for a sliver of the cost on commodity iron.

Update: Ben Galbraith responds