-
Quote
-
Login to like this post
Post
by dvgrn » May 2nd, 2018, 5:32 pm
Elementary Conduit Repeat Times
A few elementary conduits, with the naming convention used in this thread, have started showing up on the LifeWiki, mostly due to imports from the Life Lexicon.
The LifeWiki has a place to put a repeat time for each elementary conduit -- and I've been meaning to collect that information for all known elementary conduits at some point anyway, with the idea of eventually producing some kind of Hersrch For The Modern Age.
Wishful Thinking about a New Hersrch
-- It would sure be nice to be able to select an output and an input, and tell a search program "I want a stable mechanism to turn this output into that input in T ticks, and it should fit in the empty space in the current pattern, and the recovery time should be R ticks or less." The search should allow H-to-Gs and syringes, and tandem gliders (rarely useful these days but you never know) and G-to-MWSS-to-G conversions and every other reasonable-sized mechanism that we can throw into the library -- along with all the traditional Herschel conduits.
A Small Annoying Problem with Statistics
Unfortunately, the repeat time for converters turns out to not be a terribly easy thing to calculate. You can run a conduit with an R-pentomino input, for example, and then draw in another R-pentomino as soon as the sparks clear, and call that the repeat time -- but in a lot of cases there's not really any way to get an R-pentomino to that location that quickly in practice, because the just-faded sparks would have gotten in the way of the stages leading up to the R-pentomino.
The Same Thing for Herschels
Come to think of it, it really isn't easy for Herschel conduits either, in some cases, but we've gotten along with an approximation so far.
The approximation is that the input Herschel is considered to be "coming from infinity", meaning that the conduit is supposed to work if you prepend something standard to it -- usually a BFx59H. In other words, we're not doing the testing as if the Herschel just magically appears in the input position at T=0.
This really does make a difference a lot of times. If the first natural glider is going to be allowed to escape, especially, then the repeat time is higher because the usual way of evolving a Herschel will conflict with the previous outgoing glider. If it weren't for that detail, several Herschel conduits would have a lower repeat time.
Hypothetical Hiccups
This means that if a new conduit ever shows up that creates a Herschel really quickly, and/or from a different angle than the BFx59H, then we'll have to reduce the ratings of a bunch of existing conduits. Not that that would be so terrible, but -- ugh.
However, until a conduit like that actually shows up, mechanisms with a nominally faster recovery are really going to be useless almost all of the time if they can't recover that quickly when connected to BFx59H. So repeat times measured in a BFx59H context are really very usable in practice.
There's More Trouble If It Ain't A Herschel
The problem is a bit worse for converters in general, because nobody even really knows what the "default" conduit is for supplying, say, a B-heptomino to the BNE14T30, or (better example) an R-pentomino to the RF28B. Quite possibly future conduits will show up that will do a better job of making those deliveries quickly and staying out of the way.
"Instant Appearance" Recovery Time -- a lower bound
So... what should I put down for the recovery time of all these converters from the new Life Lexicon? What I've done so far is to record the IA "instant appearance" recovery time, as I mentioned above. The IA-recovery is the lowest N where an input X can magically appear at the input position at T=0 and then again at T=N ticks (or any number greater than N), and the conduit reliably will work and recover... twice.
I'm not entirely happy with this method, because it gives an unreasonably small number for the RF28B recovery time. If we use IA-recovery numbers everywhere, it seems like it will turn out sometimes that you'll combine two conduits X and Y, with recovery times rX and rY, and you'll find that the recovery time rZ of the combined mechanism will be *higher* than either rX or rY, instead of just trivially equal to max(rX, rY).
(Just like in the Herschel case, it will quite often happen that the slowest-recovering part of the mechanism is right at the connection point between the two joined conduits -- and the IA-recovery measurement doesn't test that connection point.)
I Can't Think of Anything Better -- Can You?
But I don't see what to do about it. If I use the current "best known connection" to feed in a converter's input, and measure the recovery time based on that, then we'll likely end up with the opposite problem -- someone will come up with a new conduit X, and now the combination of X and Y will have a recovery rate *lower* than either rX or rY, because rY's recovery rating is outdated as soon as conduit X is invented.
We're safe from this second problem, as long as I use the "instant appearance" method to calculate recovery times. So I might just have to stick with that -- unless anyone has other suggestions. (?)