No, there aren't any detailed designs out there that I know of. That would be the next step after this Cordership-gun proof of concept.simsim314 wrote:@dvgrn neat! Speaking of which - how would armless self replicating spaceship be built today, do we have a satisfying design?
Extrapolating from the four loop guns in the proof of concept, it seems as if an armless-constructor spaceship would either need to have four "channels", one for each loop gun -- just like the Gemini had twelve channels -- or it would need to have some way of encoding those four streams of gliders into two streams or one stream, to make the data easier to duplicate during the copying phase. For example, we could pull a trick like the one used in the 10hd and 0hd Demonoids, storing the four streams in series and bouncing them back and forth between two constructors.
Or maybe we could get away with only two streams instead of four: it should be possible to use two consecutive gliders to store 1) the choice of output glider color, and 2) the timing of the output glider. It would take a bit more circuitry to decode an encoding like that, but then again these slow glider pair constructions make circuitry much cheaper to build, so maybe it doesn't matter too much.
With two streams, storing them in series would imply a switching problem no more difficult than the one in the early Demonoids: basically you'd want to turn off the construction-arm output half the time, as the wrong half of the glider stream passed through the wrong half of the constructor mechanism.
With encoded streams, though, it's easy to lose all the efficiency you gain with glider-pair construction: you need at least twice as much circuitry to run two coordinated guns, and the coordination means that one gun often has to be delayed to synchronize with the other one. Maybe we'll come out ahead by adopting chris_c's idea for a more efficient single-channel library with 62-tick or 69-tick repeat time, but we won't really know until we try.