Elk Editor
Ubik has too many interwoven ideas to change quickly. That was a mistake. This is an attempt to fix that mistake.
Motivation
I’m interested in communication languages. Communication languages are unlike standard programming languages in that they’re uninterested in computation as such, but are instead concerned with how a network of computing processes interact with each other.
The underlying computation language for this project is clojure(script). Specifically a very pure subset of clojure. Computations are pure functions. No side effects whatsoever are allowed in the computation code. All side effects — this includes disk I/O, network I/O, changing / reading refs (which is just memory I/O), generating random numbers, state, etc., etc., ad nauseum — are handled as communication at a different level from computation.
Model
N.B.: this is just soundboarding and shouldn’t be taken as too seriously.
Computation is a mathematical abstraction. Computations don’t take time or do anything, they just are. (gcd 30 12) = 6 no matter how you determine it and no matter how long that algorithm takes. Algorithms are important, but at the level of communication, they’re irrelevant, only values matter.
That said, algorithms can have a strong effect on timing, and so play a role in the dynamics of a system, but more on that later.
A computational unit receives messages from multiple channels and dispatches to different (pure) functions depending on the channel the message arrives on.
These message processing fns receive a message and the current state of the unit, and return a new state along with zero or more messages.
Units have no control over, or indeed knowledge, of the topology of the system. They don’t know who is sending them messages, they only know the value of the message and the channel on which it arrives. Similarly, they have no knowledge of who will receive the messages they emit.
At first glance this seems to make it impossible to do anything interesting. But that’s because we habitually confound computation and coordination.
The goal of this system is to provide an interface and a language for connecting such dumb but simple and mathematically tractable units into complex systems which do anything one would expect a program to be able to do.
What’s the point of this exercise? My hypothesis is that such a system will be easier to understand, test, and modify than traditional systems. We’ll see.
License
Copyright © 2019 Thomas Getgood
Distributed under the Eclipse Public License either version 1.0 or (at your option) any later version.