Ryan Hiebert

I've been reading the code for the past few hours to see if I can implement this feature. I've also been reading Celery's code to figure out how they are...

As I've read the implementation, it's become clear that, strictly speaking, the result middleware isn't required in order to make the task pipeline work. Celery absolutely requires the result storage...

It does look like at least RabbitMQ is capable of setting task and queue TTLs. https://www.rabbitmq.com/ttl.html#queue-ttl

> You can't get an accurate queue size from RMQ via the AMQP protocol I don't think I technically need to get the size, I only need to check if...

@Bogdanp : With #134 completed, how do you envision this communicating with the rate limiter backend, so it can instantiate the barrier? Should the Pipeline middleware take the rate limiter...

The experiment I'm currently working on is to add a new middleware, `Groups`, that knows about `Pipelines`, but doesn't attempt to do anything that `Pipelines` does for itself, so that...

Unfortunately, I think that groups and pipelines are just too intertwined to attempt to keep them in separate middlewares. And, unfortunately, messages as well. I'm working on an alternative design...

I've been finding this to be a uniquely challenging problem to solve. I am making some progress. My current iteration is combining pipeline and groups into a single middleware as...

I'm not working on it anymore. I'm so frustrated with Celery's signatures that I'm questioning whether the design is anything more than a footgun. I've been investigation entirely other options,...

I agree that the links to nested resources should be from elsewhere than the api root.