Kinto VS baseplate.py

Mozilla team to https://github.com/Cornices/cornice , https://github.com/Kinto/kinto etc

Almost any download link on the internet (for example an attachment to any Wordpress blog post) could serve as an example. Since we are on a programming subreddit let's use Github as an example. When you open a repository and click the "Download ZIP" button it's a link straight to an URL like this one: https://github.com/reddit/baseplate.py/archive/refs/heads/develop.zip (this particular one is for Reddit's Python library).

Even Reddit has python backends https://github.com/reddit/baseplate.py based on Pyramid. They also have a go one. https://github.com/reddit/baseplate.go

there's a multitude of services in reddit's architecture. as far as i can tell, they mostly using reddit's baseplate framework (which has implementations in both python and go).

Finally, the problem that we didn’t experience directly, but it was mentioned during consultations with another team that had experience with pgBouncer: the Baseplate.py framework that both of us are using sometimes leaked the connections, leaving them open after the request, but not returning them back into the pool.

Also reddit team using pyramid for services https://github.com/reddit/baseplate.py

Reddit is built on neither Flask nor Django. The old monolith predates Flask and Django and is built on its own framework. Our new microservices are built on our Baseplate.py framework.

Minsky is an internal baseplate.py (Reddit’s python web services framework) thrift service owned by Reddit’s Machine Learning team that serves data or derivations of data related to content relevance heuristics — such as similarity between subreddits, a subreddits topic or a users propensity for a given subreddit — from various data stores such as Cassandra or in process caches. Clients of Minsky use this data to improve Redditor’s experiences with the most relevant content. Over the last few years a set of new ML capabilities, referred to as Gazette, were built into Minsky. Gazette is responsible for serving ML model inferences for personalization tasks along with configuration based schema resolution and feature fetching / transformation.

Baseplate is implemented in Python and Go, and although they share the same main functionality, smaller features differ between the two. One such feature that was previously on the Go implementation but not Python was deadline budget propagation, which passes on the remaining timeout available from the initial client request all the way through the server and any other requests that may follow. The lack of this feature in Baseplate.py meant that many resources were being wasted by servers doing unnecessary work, despite clients no longer awaiting their response due to timeout.

In order to make this solution work, you’ll need a web stack that can handle many concurrent requests. Reddit’s stack for most microservices is Python 3, Baseplate, and gevent. Django/Flask also work well when run with gevent. gevent is a Python library that transparently enables your microservice to handle high concurrency and I/O without requiring changes to your code. It is the secret sauce that allows you to run tens of thousands of pseudo-threads called greenlets (one per concurrent request) on a small number of instances. It allows for threads handling concurrent duplicate requests to be enqueued while waiting to acquire the lock, and then for those queues to be drained as threads acquire the lock and execute serially, all without exhausting the thread pool.