ConwayLife.com

The reverse caber tosser may be all good in theory, but has anyone proved it can create everything, and clean up its own junk? I'd like to see a pattern that uses a RCT to create (say) a pulsar and nothing else. Most of the designs create inordinate amounts of junk, so as a start point we could try to create a pattern that clears all the junk of a GPSE.

erictom333 wrote: March 17th, 2022, 3:50 am The reverse caber tosser may be all good in theory, but has anyone proved it can create everything, and clean up its own junk? I'd like to see a pattern that uses a RCT to create (say) a pulsar and nothing else. Most of the designs create inordinate amounts of junk, so as a start point we could try to create a pattern that clears all the junk of a GPSE.

erictom333 wrote: March 17th, 2022, 3:50 am The reverse caber tosser may be all good in theory, but has anyone proved it can create everything, and clean up its own junk?

We have no proof yet. We have a pretty good proof outline though. There are just a lot of details to be filled in, and those details haven't been engineered.
There could be possible unforeseen problems in the design that show up during that part.

erictom333 wrote: March 17th, 2022, 3:50 am I'd like to see a pattern that uses a RCT to create (say) a pulsar and nothing else.

You are not the only one. The problem is, the current 17-RCT design has a lot of layers to transform construction arm's complexity into time complexity. And even for 329-RCT there are still a lot of things to do to make nothing but predefined object.

erictom333 wrote: March 17th, 2022, 3:50 am Most of the designs create inordinate amounts of junk, so as a start point we could try to create a pattern that clears all the junk of a GPSE.

Yep. The junk can be split into 3 parts: The junk at the construction arm's location, the junk at the GPSE's starting points, and the GPSE's ash.
The first one can be cleaned (almost) trivially, the second one can be carefully cleaned by firing non-intersecting gliders to the start point, and the third one can be cleaned up by enormous fleet of corderships (or other sparky diagonal ships of which 2-engine cordership is probably the simplest).

Let's do some rough calculations. There are 12 different objects in GPSE's ash. There are 5 GPSEs in the 17-RCT design. That's 60 corderships to construct. Each cordership costs about 100 slow gliders from single block. That's about 6000 slow gliders total just for cordership construction.
And these slow gliders are yet to be encoded in the initial binary number. Even if we assume that each glider costs 20 bits (that assumes a reasonably efficient construction arm, with different move and fire operations), the overall size would be at least 2^(10^5). I think there's no hope in running that in Golly for now with Hashlife algorithm.

And keep in mind that this figure is very underestimated.

Pavgran wrote: March 17th, 2022, 7:49 am Let's do some rough calculations. There are 12 different objects in GPSE's ash. There are 5 GPSEs in the 17-RCT design. That's 60 corderships to construct. Each cordership costs about 100 slow gliders from single block. That's about 6000 slow gliders total just for cordership construction.
And these slow gliders are yet to be encoded in the initial binary number. Even if we assume that each glider costs 20 bits (that assumes a reasonably efficient construction arm, with different move and fire operations), the overall size would be at least 2^(10^5). I think there's no hope in running that in Golly for now with Hashlife algorithm.

And keep in mind that this figure is very underestimated.

Yup -- for example, the 60-ish Corderships will all have to be launched after the last glider comes in and the incoming GPSEs crash into (something unknown but fairly easy to design, that was constructed along with the target pattern to catch the GPSEs without releasing any gliders). That means that what we really have to build is a seed for those 60-ish Corderships, which multiplies the cost by another factor of ten or so.

Then there might not be room in the center of the GPSEs for all the seed structures that will have to be triggered simultaneously -- so instead in some cases we might have to build a "meteor shower" seed off to the side somewhere, out of the way, that produces a slow salvo of gliders that construct the 60-ish Corderships... or clear out more space (cleaning up the "crash site", which won't be nice simple GPSE ash anyway) and then build and trigger a seed for the Corderships -- probably another factor of ten or so.

This all makes for a big project, but it's not completely beyond current engineering capabilities -- we know how to do all of these things. We could build a series of demo patterns that are runnable in Golly, each of which would demonstrate one of the construction or cleanup tasks.

Pavgran wrote: March 17th, 2022, 7:49 am You are not the only one.

Yeah, this question keeps reappearing. Here's a post from the last time this came up, just over a month ago. that links to a painfully detailed response from one of the earlier rounds of this discussion.

Long story short, it really seems extraordinarily unlikely that any unsolvable engineering problems would come up, if the community decided to focus on solving all the unsolved pieces. Unfortunately it's not a very interesting project for most people, because putting it all together and watching it run just plain won't be a lot of fun.

It should be possible to build an example pattern which contains -- already present at time 0 -- all but the last few gliders aimed at the construction arm (these last few gliders -- 30 or so -- will be generated by the actual RCT mechanism). That way, the bounding box and runtime will only be linear in the recipe length, but it would be able to show off all of the construction and the cleanup mechanism.