I’d seen that before, but am glad to have the confirmation that it is not possible to add Ember add-ons until that upgrade is done. So it sounds like adding ember add-ons will be possible soon–once the upgrade is complete. And that sounds great.
I think that’s an interesting question. Here’s my two cents:
In terms of using the discourse “abstracted stuff” v ember add-ons: I could be wrong, but I think that using ember add-ons for specific tasks in plugins might be easier to maintain, in the cases where you are trying to do something different than what Discourse is doing already. This is my thinking:
Example here is wanting to add a brand new drop-down in a plugin. That distinction is probably important–I’m talking here about trying to do new things in a plugin that are not done in the discourse code-base, and the question is whether to start with discourse methods or a separate add-on.
A lot of times you really don’t have a choice. For example, if I wanted to add a custom field to topics, I’d always customize on top of discourse’s pre-built methods and code.
But If its a targeted piece of functionality–like a drop-down for a new purpose–I’d already be in the case where, if I use the discourse methods, I’d be adapting them to things they weren’t targeted towards.
Option 1: I could try to take the select-kit code I see in, say, category-chooser, and insert it in a new spot (that’s not about categories), and then try to customize it to be about what I want the dropdown to be about, instead of categories. This is the task that I described earlier as being tricky.
And it could be difficult to maintain–because if the discourse team changes something in how the category-chooser select-kit code works, then that could change my new customized drop-down, but in ways that could be surprising (bc I had customized it so it works slightly differently than the actual category-choose dropdown).
Option 2: I could insert something from ember that is robust but that is also made to be customized, where I can see fairly clearly how the code actually works. In that case, I might miss out cool new features that discourse might add to its drop-downs, but I would be able to stay on top of how my drop-down works more easily. So this is probably the best option, I think, if it is possible.
Option 3: Code it entirely from scratch. That’s where I’ve been tending to end up. When the coding is done, it’s nice to have code that I fully understand and can customize. But of course it takes longer, and (the initial versions at least) are unlikely to be as powerful and robust as what the discourse team or the ember team have built.