Dan Allan
Dan Allan
in #383 we added a boolean switch for (2), `reject_undeclared specs`. By default, we are permissive, i.e. `reject_undeclared_specs: false`.
Yeah that article has a lot in it, it took me a couple passes to follow it. Specifically I think the lesson for us is, asyncio makes it easy to...
#478 enables this kind of thing. We should address this timeout situation directly, at least with a better error message recommending tuning `batch_size`.
That is a useful perspective and something I think I have also heard our scientists ask for: a way to attach to the queue with something that feels like RE...
I should add, to complete the thought, that this leads me to suspect we may want both: 1. A REPL client to the queue, with all the benefits of _intermediated_...
Our current answer for (1) is of course: https://github.com/bluesky/bluesky-queueserver-api
Thanks @callumforrester. Since Diamond’s use case is distinct, I think if we can engage with it in this design process we will significantly improve our chances of choosing a strong...
Thanks, that is all very helpful, and it makes sense to me. Even for our simple use cases, I can imagine wanting to sanity check, "Do these names exist in...
An unrelated observation: the [Jupyter protocol](https://jupyter-protocol.readthedocs.io/en/latest/messaging.html) uses five 0MQ channels, each for a different function (heartbeat, control, shell, stdin, IOpub). I believe queuesever now has two: the document steam and...
I don’t think a queueserver-api shim would address the costs risks and costs of making that discontinuous jump from the current in-process operation to the queueserver-based operation. The primary goal...