Mycroftiv dev notes


The goal here is to create a generalized hypercubic platform for distributed processing and applications. The core primitive of setting up a hypercubic topology and routing messages around is accomplished with the current tools, and whatsit.rc can make and control smallish hypercubes. it can also do some predefined slices of larger hypercubes.

There needs to be a library of 'hypercubic primitives' such as have already been prototyped - the generalized shuffle exchange template, fanout, and coalesce, along with some others. What MPI offers is a good reference for ideas.

To make a 'pluggable cube' where nodes can attach and detach at will on the fly is a midterm goal.


The mrsulu.rc framework is fun and works and at least one substantial application (the moon sim) should be created within it. It should also have at least one blog post written, and this would probably be a good time to try to actually make contact with some of the ncube people or other researchers.

A real Time Warp implementation in c using hubfs as a core (or perhaps burnzez work in progress?) will be one of the next steps. The goal is something less ambitious than what TWOS was doing, but still more generalized and os-like than a pure sim platform such as ross.

Long term, the dream of a fully reversible operating system is worth pursuing. Such a thing would probably need to be a fundamentally new os, because issues of reversibility have to be considered at every single design point.


The near term goal is historical preservation of the original c code in a form that is easier to build than the current Watcom project files and will compile and run in any standard unix derived os including plan 9 ape (with perhaps extra glue needed for ape). The medium term goal is a port of the core engine to a form where it can be used as a pluggable sim-backend within the distributed time warp framework. In other words, each time warp object can have its own attached private IMP which it can use to simulate a physical system. IMP is easily reversible in the manner needed by Time Warp.

Longer term, finding possibiities to distribute from within IMP should be pursued. Matrix multiplications of large enough matrices are suitable for parallelization, and the mechanism of decomposing a physical system into subsystems has obvious relevance.


The first ANTS 5.x release should be soon, including Burnzez new hubfs and spawngrid tooling (undocumented) and hypercubic/timewarp scripting (semidocumented).




Maintenance but should really get some more dev time and effort.

Rabbit Hole