Tom Nicholas

> This feels like the right path forward to get substantial improvements for deep trees. It might be, but we need a much clearer idea of what is actually slow...

> Max pointed out that such accessors can be useful for adding state without the need to subclass FYI they don't really add state in a reliable way anyway, see...

> The lacking code completion (at least when that is not built on top of type hints) is a bug in xarray: the accessors are not included in __dir__. Yeah...

> it may be worth not including as a motivation or even explicitly discouraging in the xarray docs with an alternative recommendation for how to build stateful extensions. Totally fair...

@chuckwondo, all of these criticisms are valid (and I agree with a several of them), but they are criticisms of a design pattern that is well-established in xarray (and as...

> wearing out my welcome Not at all, I'm pleased to see the thoughtful and constructive engagement! However, I must say I feel like this question in particular has become...

I don't know much about the details of the pydap side but I also like the idea of working towards having a single source of truth parser, and that parser...

Hi @mraspaud - thanks for this contribution! I can see how this might be useful. I apologise for the indeterminate state of datatree right now. > From what I understand,...

As there have been multiple relevant changes upstream, it's not clear if this issue is still relevant. @mraspaud can you please try it with the upstream code (including the `open_groups`...

Thanks for the quick response @JosepSampe ! > When we began developing the cloud proxy internally in 2019, the framework you mentioned didn't exist. That's totally reasonable, but I'm not...