# IRC log for #marpa, 2014-12-17

All times shown according to UTC.

Time Nick Message
00:32 jeffreykegler joined #marpa
04:27 jeffreykegler joined #marpa
04:45 ronsavage joined #marpa
04:47 jeffreykegler https://github.com/jeffreykegler/kol​los/blob/master/notes/misc/loup2.md
04:47 jeffreykegler Just pushed a new version of the note on Loup's empty rule tutorial.
04:48 jeffreykegler While the proof that the algorithm is fast is fairly easy, it occurred to me it is not trivially obvious that my algorithm is correct -- that it does what it claims to do.
04:49 jeffreykegler It's been running a lot, so if it was wrong we'd probably know and I must have felt certain at the time I first coded it that it was correct, or else I would have included the proof in the marpa.pdf file ...
04:49 jeffreykegler but looking back after this time, it is no longer as obvious as it once was ...
04:50 jeffreykegler so I've added a correctness proof.
04:50 jeffreykegler Usually programmers know this sort of algorithm is correct, because they found it in a book, :-)
04:50 jeffreykegler but in this case we are writing the book.
04:56 btyler joined #marpa
06:36 rns joined #marpa
06:37 rns jdurand: re http://irclog.perlgeek.de/​marpa/2014-12-16#i_9815905 — :) great, thanks.
06:38 rns jeffreykegler: re http://irclog.perlgeek.de/​marpa/2014-12-16#i_9815931 — my thoughts exactly.
06:38 rns left #marpa
11:33 flaviu joined #marpa
13:42 sirdancealot joined #marpa
13:42 slothmachine2 joined #marpa
13:43 lwa joined #marpa
15:20 jeffreykegler joined #marpa
16:10 hobbified joined #marpa
17:45 jdurand_ joined #marpa
18:10 flaviu joined #marpa
21:22 sirdancealot joined #marpa
23:11 flaviu1 joined #marpa
23:22 jeffreykegler joined #marpa
23:23 jeffreykegler I've been exchanging emails with Loup Vaillant, who's doing the Earley's tutorials, and we're discussing sharing code.
23:24 jeffreykegler It seems natural, since we both use Lua, and we'll need a lot of the same routines.
23:25 jeffreykegler It's important to emphasize, this is *not* about combining the two projects -- just sharing code as the opportunity presents itself.
23:25 jeffreykegler Loup is focused on building the thing up from first principles, understanding it as he goes and, when in doubt, keeping things simple.
23:27 jeffreykegler With Marpa, my idea was that to break through the indifference barrier, there would have to be a useable implementation, and so from the start I've emphasized optimization and scalability, and have been willing to pay the price in increased code complexity.
23:29 jeffreykegler So my ideas is that we can work on pieces of Kollos, have them become part of Loup's code (if only as alternatives to simpler implementations), and then "borrow" them back when assembling Kollos.
23:29 jeffreykegler The advantage for Kollos, to my mind, is that they'll be easier to find and understand if they're part of Loup's tutorial, and they'll get some exposure and perhaps testing.
23:34 ronsavage Sounds like it could be quite a productive initiative. As for 'I've emphasized optimization and scalability' I agree that should have been your goal. It can be tempting to just get something going and then optimize it later, but I get the impression you went all out for that right from the start.
23:35 jeffreykegler Yes.  Folks have muttered stuff to me about "premature optimization" ...
23:35 jeffreykegler but I tell them that, except for speed, this problem was solved in the 1960'
23:36 jeffreykegler s.  Waiting 50 years to optimize really can't be said to be doing it prematurely. :-)