Rally for PHP-FPM

[otobio]

Description

What is the future plan for PHP-FPM? Currently, this technology appears to be falling behind the modern web's demands. While PHP-FPM (FastCGI Process Manager) has been a reliable tool for managing PHP processes, it lacks advanced features that could enhance web performance, such as early hints. This stagnation has relied on just pool of process management and really nothing else.

If the web is PHP's strength, it should lead the way. Unfortunately, PHP-FPM is falling short in this regard, as its architecture lacks modern optimizations crucial for faster loading times and enhanced user experiences. While other components of PHP, like interpreter performance and JIT, are evolving, PHP-FPM remains stagnant. It is arguably one of the most critical extensions, and I hope we can rally a community of passionate developers to drive its improvement, just as we have seen with other extensions - It can't be an after thought solution.

PHP-FPM must move fast and at-least at the pace of the top web servers and adopt new technologies and methodologies. Integrating features like HTTP/3, enhanced caching, and improved concurrency models will be crucial. By embracing these advancements, PHP-FPM can boost performance and equip developers with the tools needed for efficient web applications.

[bukka]

The primary focus was mainly on stability and getting the bugs fixed. There has been a significant improvements in the test coverage. FPM has its own test framework and also a new tool called wst has been developed for integration tests with the server with some initial tests in the separate project wst-php-fpm. Many bugs also got fixed and we are at the state that we have just few open bugs. This is a big difference compare to years back where we have many open bugs and no tests. It means we now have some confidence to do fixes without introducing new bugs.

In terms of priorities, it is to get the last unfixed bugs fixed which will require some significant improvements to the testing tool as especieally the ondemand scaling issues require a good coverage. So the priority is still the stability which is also something that is very important for most users.

There are bunch of things that could come after that in terms of features - you can see my TODO list in https://github.com/bukka/php-todo/blob/master/php-fpm.md . This is ordered by priority in each section so you might get some idea. Of course, if someone implements some of the features quicker, then it gets priority for the review.

From the bigger features, we could look first to the worker mode that does not seem too dificult. There is also an idea of the pool manager which would resolve many existing issues but that's a bit harder and will take time. That could also allow using TLS and possibly other protocols. Early hints are a bit tricky because they require support on the server for fcgi.

I think if you really want to use those features quickly, you should look into FrankenPHP as it offers all those mentioned features already. The plans are actually to recommend FrankenPHP for users that want to use those features and make it more official PHP project so people are not afraid to use it. That doesn't mean that FPM will get deprecated. It's just that it's more for users that require stability and don't need those extra features.

There is not much to do in terms of this issue (all is already known and you can see links on existing feature requests in my TODO list) so I'm going to close this.

[bukka]

Just a note that I closed as Not planned, because it's not completed. It should be more duplicate but there is no single issue for this.

In any please send those kind of things to the mailing list as it's our place for discussion - issues are just for bug reports and specific feature requests.