Investigate shaving milliseconds off target language binary start times

[Marak]

While the current speed of spawning and executing the target binaries for every language ( like node or php ) is very fast, I'm pretty sure we can shave off a few milliseconds from each one.

The idea is that we don't need to start a full-featured instance of node or php. We only need a specific set of modules / apis which we support. While most of the features we will require, it may be possible in some languages to do a custom build of the binary that reduces start-up time execution by a few milliseconds.

Very low priority.

If anyone has suggestions or any specific language binary please let me know.

[pyhedgehog]

What about pool of prefork processes? It could be easier to implement. It will be more compatible. It's better scaled to new languages.

[Marak]

Not sure what you mean by that.

[pyhedgehog]

Something like this: http://cx-oracle.readthedocs.org/en/latest/session_pool.html