>>I was refering to the case where we have to code for two events in separate reference frames
>
>
>But if your program has code for separate reference frames then you aren't using absolute space as the co-ordinate system.

My assumption was that your code is the observer, i.e. it necessarily works in some local coordinates, and then the master code has to translate each observer into absolute space. If I got you right, your assumption was that we're already having each observer code in absolute space - which is actually the same in the end. The code that would need to convert local coordinates into absolute would just be run at a different time. In my case, it would come into play when the individual observers are aggregated, in your case, immediately for each of them.

>But I think observed effects like time dilation and length contraction need to be accounted for in the model. So that's a problem that would need solving.

Exactly - that's what I'm saying. The transformation of local coordinates into absolute space wouldn't be linear. It wouldn't even be constant. The functions involved may take time and velocity as parameters - I don't know that, and surely don't know who does. My last physics class on that matter was about 30 years ago, and I actually think of that more in terms of Minkowsky space or a differential manifold, where the metric isn't necessarily just Pythagora's hypothenuse measure, it may well be a tensor.

>Can you model motion for all observers and still account for time dilation and length contraction?

That's the point - while I can imagine a metric tensor for such a space, the question which functions are its elements, and what are their variables, it's not math anymore, it's physics.