Mark Hammond

We briefly discussed UDL yesterday, so it's worth pointing out that the fxa-client crate has [extensive docs in UDL](https://github.com/mozilla/application-services/blob/main/components/fxa-client/src/fxa_client.udl) which we should make sure are rendered. I believe some edits...

In #2264 I wasn't as focused on the exact result of `isinstance` etc, more about the error. > We always generate an ABC for trait interfaces -- regardless of with_foreign....

I think I'm suggesting that: > In all cases, isinstance can reliably check if some object implements a trait interface. `isinstance` probably doesn't matter as much as pyright etc being...

yeah, that's horrible :) `typing` is kinda counter-intuitive to me - it's not a runtime thing. Trying to derive from their ABC works at runtime and fails at typecheck time....

@bendk I think you fixed all this as part of the pipeline - should we close this? (aside: I really don't like `type_name` in the Python bindings - it's part...

The last release of askama was over 18 months ago, so this might make sense. The changes you list between the forks all seems improvements to me and rustdocs moving...

Thanks for those changes, but do you mind *adding* an entry for this?

It looks like the only new dep here is [once_map](https://github.com/a1phyr/once_map) which looks OK and I'd be fine approving. I did strike a strange issue when moz-central insisted on pulling 0.4.20...

> rustc-hash v2 has a new algorithm compared to v1 A quick look shows that it's really the use of `FxBuildHasher` which would avoid trivially using either version.

Just got confirmation the rustc-hash version isn't going to be a problem for this, so I'm fine with landing it once updated!