Eric Wieser

Regarding naming; we could have the function named `conj`, but still call the typeclass `star_monoid` etc. @semorrison, independent of whether we go the bundled approach, what's your view on star...

I have a slight preference for B, but also I'd never seen the "star" terminology before this so am not particularly equipped to see the downsides in B. I don't...

Do we actually need our entrypoints to be specific to debugging, or would any plugin system with setup/teardown phases do the job?

> EDIT: Thre are probably better/newer examples. If we go down that route, my vote would be for `pluggy`, since that's the plugin manager used by `pytest` and `tox`. It...

> There is #1225. This could fit into a more robust startup customization, but I haven't thought much about exactly what that looked like. I think we should treat this...

@garmin-mjames: Note in principle that cocotb could support only `entry_points`, and we could provide a `cocotb-plugin-hacks` package that provides an entry point that reads `COCOTB_STARTUP_ROUTINES` (or vice versa)

https://github.com/commonmark/commonmark-spec/pull/745 is related

@mdickinson: any chance you're willing to review one more buffer-related PR? This one isn't quite as old as the other two of mine you rescued, but it's getting close!

I'm afraid I don't remember what I considered at the time (which is obviously a lesson in writing better commit messages). I'll see if I can come up with an...

I suspect consistency with numpy was the main argument; along with consistency with `to_list`: ```python >>> import numpy as np >>> arr_0d = np.array(42) >>> mem_0d = memoryview(arr_0d) >>> len(arr_0d)...