Mleku, we're dev friends. You don't need an excuse, to talk with me about our pet hobby.
I just disagree. I see a point of the kind number, but also the m/M tags (the AI automatically understands the purpose of the sample events, when they're in the json, for example).
Kind includes use case information and you'd have to find some way to sub for that, but it is best-defined with a key, so that it can be independently and asyncronously defined without a central authority.
see, that's my point
a kind number needs a registry more than a descriptive tag name. a P tag would be perfect to signify protocol. you don't need permission to put "P":"stella's awesome protocol" and nobody is gonna fight with you about that (i'm just being silly regarding the actual name but the point holds).
kind numbers are supposed to signify protocols so when you look at what people have designed, there is protocols with multiple tags and protocols with single, ever increasing varieties of tags. the semantics of tags and kinds are clearly intertwined. kinds are just a kind of tag, IMO.
it has lost its meaning because nobody defined one. it is indexed, it can be filtered the same way as a single letter tag, and its semantics are absolutely intolerably ambiguous, and not human readable.
dealing with making a reasonable index of them in my codebase was one of the most tedious tasks of the whole thing, second only to debugging.
