Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2016-01-05

| Channels | #marpa index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:09 ronsavage VsyachePuz: JK Will have to give me the link.
00:13 Idiosyncrat joined #marpa
00:13 Idiosyncrat Scott's SPPF paper: http://dinhe.net/~aredridel/.notmine/PDFs/Parsing/SCOTT,%20Elizabeth%20-%20SPPF-Style%20Parsing%20From%20Earley%20Recognizers.pdf
00:15 Idiosyncrat Marpa's parse bocage is the same as Scott's SPPF's -- there are not two distinct theoretical structures whose equivalence needs to be proved.
00:16 VsyachePuz Idiosyncrat: but it is better to describe bocates in their own names by you own words (as in that tutorial about earley)
00:17 Idiosyncrat huh?
00:17 Idiosyncrat You mean rewrite Scott's paper using Marpa style terminology?
00:19 VsyachePuz yes, in Loup Vaillant -like format
00:19 Idiosyncrat You liked Loup's work?
00:19 VsyachePuz yes, it is almost perfect according to my requirements
00:20 Idiosyncrat It has received much praise.
00:20 Idiosyncrat I wrote him some time ago "complaining" about how much better people liked his tutorial style than mine. :-)
00:20 VsyachePuz it should have be a bit more linked, but that just requires more work
00:20 Idiosyncrat He was amused.
00:21 Idiosyncrat Perhaps Loup will someday describe SPPF's -- his tutorial series is proceeding slowly.
00:23 VsyachePuz as I understood, he achieved his goal (ho receive something practical), and have no future motivation
00:26 Idiosyncrat Yes, in my work, there are a lot of write-ups and explanations that I would have liked ...
00:26 Idiosyncrat and I either had to create them or do without.
00:26 Idiosyncrat maybe 10 years from now all this stuff will exist, ...
00:27 Idiosyncrat but then the field won't be so exciting. :-)
00:27 Idiosyncrat Meanwhile, I'm writing up the Leo algorithm, as it is used in the context of Marpa, and that is proceeding, but slowly.
00:27 VsyachePuz Idiosyncrat: it is good that there is many explainations, but such explainations should be moved to separate pages from main path (and linked)
00:31 VsyachePuz Ideally i want to see: 1) historical timeline of theory developement (should name features) 2) page for each feature, why it is important 3) linking all there to implementation. This will allow to claim that parser implements all existions scientific achievements, and will allow to explain how each part of parser work exactly and why (so, it's a huge amount of text which should be better linked than a book)
00:39 VsyachePuz For now i see, that timeline contain the term "SPPF" but it is not mapped to implementations
01:00 Idiosyncrat Re the mapping to implementations, that's a kind of infinite regress.
01:01 Idiosyncrat That is, every implementated algorithm, the proofs proceed via a theoretical model and the belief that the implementation matches the theoretical model ...
01:02 Idiosyncrat is not formally justified, because all that would do is create another formalism to match to the implementation.
01:02 Idiosyncrat So for example, I don't think there is any proof that ANTLR's code matches the theoretical model in the papers.
01:11 VsyachePuz let's rename "regress" into "requirements tracking"
01:15 VsyachePuz the requirement "implements top scientific achievements" are divided into "implements SPPF". I am unable to track it the last requirement is implemented, or not. Or how to implement it. Loup describes how to implement the first part, but his implementation doesn't contain SPPF (wikipedia have good short section)
01:17 VsyachePuz https://en.wikipedia.org/wiki/Talk:Earley_parser#Constructing_the_parse_forest
01:18 VsyachePuz Loup describes first two paragraphs, and the rest is unexplained
01:22 VsyachePuz i believe that chinese will implement it first (as japanise did in 2004). This is because they gather all this scattered information into linked massive during translation
01:44 VsyachePuz "some parts of the algorithm (particularly the construction of the parse forest) don't seem to have good descriptions which are easily accessible to non-experts" (q) http://joshuagrams.github.io/pep/
02:12 kaare_ joined #marpa
03:46 ronsavage1 joined #marpa
08:37 ronsavage joined #marpa
09:38 pczarn joined #marpa
12:40 pczarn joined #marpa
13:21 kaare_ joined #marpa
13:46 VsyachePuz Peter Naur (letter N in BNF) RIP at age 87, 2016-01-03, https://www.facebook.com/henglein/posts/1253954421288286
17:32 ceridwen joined #marpa
20:42 ronsavage joined #marpa
20:43 ronsavage Sad to hear about Peter Naur.
22:29 ronsavage JK: I've just realized: If all the programmers and libraries who ought to have a copy of your upcoming book bought (sic) a copy, then the royalties would be a big help to you.

| Channels | #marpa index | Today | | Search | Google Search | Plain-Text | summary