Joep Meindertsma

Creating a new datatype or not is a tough decision. I tend to gravitate to keep the amount of datatypes to a minumum. If a datatype can actually be represented...

@jonassmedegaard Agreed, that's a more sensible approach. I've updated the issue.

> Why do messages need to be tracked in (virtually endless) arrays? I don't think they _can_ be arrays, because of performance. You can't serialize all messages when someone wants...

I'm not trying to reinvent matrix, I'm trying to find a way to model 1-n relationships in atomic data. Chat messages are just one use case. And events are a...

Looks interesting, thanks!

> This is essentially a generalization of the model you initially suggested -- with nextPage / previousPage you're just implementing a LinkedList of arrays. That's true, in a way, but...

> I see two approaches that directly mention Collections: "Add a new DataType for Collections (note: current favorite)" and "Allowing Collection URLs where ResourceArrays are used". What's the difference between...

It will be an extremely small wrapper around functions that already exist in `atomic_lib`, so it will be part of that library. But I can also add it as a...

Thanks! It also make sense to keep Git management in consideration here, since running these checks on commit would prevent issues #97

> For in-browser validating the ideal would again be to not depend on network access, by implementing a validator in JavaScript and shipping that both as an NPM library and...