Jacques Carette

I think we need more precision about what the typing rules mean. For example, we should start with a grammar for the types themselves. There's also some difficulties. For example,...

Part of the point of doing this is to really pinpoint where the problems are. And perhaps get an advanced peek at problems we hadn't yet spotted.

`ConstraintKinds` are indeed a nice piece of Haskell, and could be part of the solution. At a higher level, here is how I use the Haskell parts to connect up...

Probably a good idea. I'd never written that down before, nor even really thought about it consciously. But this issue pushed me to make explicit why I was feeling uncomfortable...

We should circle back on this. We need to understand the problem, so that we can settle on a design.

UID = "Unique IDentifier". It is a way to refer to a piece of 'data', and gives an 'identity' to data. Using strings for this was a horrible hack that...

> My first thought is that the ADT should denote the formation and possibly "change history" of the knowledge/chunks they are given to. If by "formation", you mean the context...

Note: I've disliked our use of 'chunk' for > 3 years now. When we started ~7 years ago, it made sense, since our thinking was along the lines of "literate...

While I'm quite keen on "knowledge fragment", I'm also quite aware that "knowledge" has been mis-interpreted (and it's vague). So I'm still trying to figure out good terminology. I do...

To confirm: we want UIDs to denote a single fragment. Furthermore, we probably do want something that's part of the UID to carry the information of what is actually available...