chris_c wrote:Hmmm... well the idea does seem at least vaguely sound. Here is a more complete example pattern. The boat reappears at (-3,-6) shift and a second glider from the SE is used to delete the blinker. The input NE glider is of the opposite color to the output NE glider as would be the case in the final spaceship.
I just tried dropping the three-scan LWSS recipe into this sample pattern. It seems to need gliders of the wrong color -- i.e., not the color of the gliders produced by the rakes:
Code: Select all
x = 512, y = 439, rule = LifeHistory
Can't shift the recipe over by just 1hd, because the LWSS only returns to the same phase every 2hd. Am I missing a trick here?
A possible nice thing about these reactions based on *WSSes and lateral gliders is that they can easily reach through half-bakery trails from 180-turners. If the timing works out badly for producing a glider for one of the trails, working from outside to inside, then an *WSS+glider trick could work equally well in the southeast.
I'm working on a script to build a complete HBK body, with a config file to allow the choice to send/not send a glider at each point of each scan. For starters I'm going to try getting a 24-scan HBK connected to a working suppression unit. Might 24 scans actually be enough to complete a knightship, with massively parallel construction?
I'll try to build in the option to do double/triple/quadruple-width scans, with multiple eight-half-bakery resets per scan... but changing the scan width also changes the required initial width between the A and B trails, which changes the way the suppression has to be done -- it gets a little bit complicated.
The A1:BA2C design is even more annoying than I thought for long HBKs with many resets. I guessed, somewhat at random, that the distance from A1 to B is going to have to be something like 1000fd, given the distance that the whole structure is going to drift along its length. The shooting range, however wide it is, will be very close to B in the northeast, and very close to A1 in the southwest (because the shooting range is really an exact diagonal, whereas the half-bakery trails are only almost-diagonal).
At 1000fd total A1-to-B width, a knightship could support only about 1000 / (39+8) = ~21 single resets before the available shooting range narrows down to nothing. For a double-width shooting range it would be 1000 / (39+2*8) = ~18, and so on. So we might need an even wider A1-to-B distance.
But if A1-to-B is wider, we have to make B-to-C a little wider as well, to let A2 drift between them. A1 and A2 have to move in step with each other, so that the half-bakery kickback always works. B and C stay in step with each other also, but at a slightly different angle -- they have the 39N half-bakeries for each scan, but they skip the 8N half-bakeries for each reset. So B and C drift slightly with respect to A1 and A2; you just need 8N extra lanes per reset, or about 170 free lanes between B and C for A2 to drift in, for a 1000fd A1-to-B design:
Code: Select all
x = 3538, y = 2723, rule = B3/S23
Most of these adjustments break our current suppression mechanisms, or at least require a lot of custom fixes, so I probably won't worry about suppression until we're sure we have the body design right.
I'm tempted to go back and look at the nice simple seven-glider A:BC setup again, with none of these B/A2/C complications as the A-to-B distance gets wider. Couldn't we build all three A gliders on the NW side, using MWSS+G or LWSS+G tricks, with just half a dozen resets or so?