Actually, I like the idea of 1080. It maps to degrees very well (Degrees are more familiar to most people than radians), and as he says there are more divisors. Also, 1080 directly equals 0, which can't be said for 1256 and 0.
Divisors aside, I had the following idea: adding sin and cos to the program would allow for some simple trig operations I was working on. Most functions in the program give and accept values very close to the range [-1000, 1000]. If we make the range [-1000, 1000], and the domain as close to this as possible, we can have cyclic nested cos and sin, which should help random DNA matain the nrg level of the original number.
Basically I'm trying to match up the spread of the domain and the spread of the range to the same interval. If as many sysvars as possible map to and from the range [-1000, 1000], I think we'll have a higher probability of mutations successfully combining sysvars.
In the end, there are alot of changes I want to make that break backwards compatibility. What I'll probably end up doing is building a new DNA specification that the various ideas I'm playing with can go to. Old bots will still run the same, but the nw specs will contain new features, and bots will have to include something like DNA2 at the top of their DNA files, or something like that.
I'll keep this under my hat for that.