Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2017-02-28

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

All times shown according to UTC.

Time Nick Message
00:09 chansen_ joined #marpa
00:11 iarna joined #marpa
01:04 idiosyncrat joined #marpa
01:05 idiosyncrat ronsavage: re https://irclog.perlgeek.de/marpa/2017-02-27#i_14171351 -- I hadn't seen it, but I'm very familiar with the stuff in the perlcall POD.
01:05 idiosyncrat That's how I originally implemented perl callbacks, before switching to the current event-like mechanism.
01:07 idiosyncrat ( The color scheme makes that web page very hard to read, btw. )
01:08 idiosyncrat The hardest thing about the perlcall mechanism is exceptions and errors, which must be passed up and thru C and then back into Perl.
01:10 idiosyncrat There are warnings, exceptions and fatal errors, both "external" (either caused by a application mistake or expected in some other sense) and "internal" (caused by programming errors).
01:11 idiosyncrat There's also the overlapping concerns of debugging and tracing.
01:12 idiosyncrat I actually did get all of this working using the perlcall mechanism, but then I considered having to maintain it going forward, and decided to rip it all out in favor of the current event-style mechanism.
01:13 idiosyncrat By event-style, I mean that when the lower layers want a Perl call, they pass an event up to Perl.  The Perl call is then executed at the top-level, in Perl, where normal error, exception and warning mechanisms are available.
01:13 idiosyncrat Results are collected and then passed back down.
01:16 idiosyncrat jdurand, ronsavage: Yes, it was me who brought up CMake.   All the cool kids are using it these days, it seems. :-)
01:16 idiosyncrat Seriously, Roberto Ierusalimschy (Lua) used it in one of his projects, which is always a good sign.
01:21 idiosyncrat hobbs: re https://irclog.perlgeek.de/marpa/2017-02-27#i_14171466
01:22 idiosyncrat to explain my 'tude about dependencies, an example from early on in Marpa's existence.  I had taken on a dependency, and the module switched to a newbie maintainer.
01:23 idiosyncrat Whose first release was DOA and broke everything which depended on it.
01:23 idiosyncrat The module's team was perhaps shy about all this, whatever, but for whatever reason stayed silent for a couple of weeks.
01:24 idiosyncrat The problem was of a kind which didn't clearly indicate the source, so I spent two weeks debugging a very mysterious glitch, effort I might have saved had I known more.
01:26 idiosyncrat I know when I started the Perl community was very very big on leveraging CPAN and reuse, ...
01:27 idiosyncrat and to be sure, in certain contexts, it makes sense to rely on, and leverage the community -- I certainly do so very aggressively when it comes to dependencies for Marpa's development environment.
01:27 hobbs yeah, but people are very careful not to treat toolchain stuff that way. It gets shared ownership, careful maintenance, and long discussion and lots of warnings before breaking anyone, because so much depends on it, literally :)
01:29 idiosyncrat hobbs: Module::Build was core as of 5.10, which was why I used it.  But then a new policy came in whereby core modules can be un-cored, ...
01:29 idiosyncrat which kinda contravenes the concept, because core to me meant I could rely on it being there.
01:31 idiosyncrat hobbs: So I have to admit to a certain degree of disappointment in the maintenance of the Perl toolchain.
01:34 hobbs it doesn't need to be in core for you to rely on it being there... in fact you get a more reliable guarantee if it's not, since you can require a particular version that maybe fixes an important bug, instead of just living with the fact that maybe you got the version from 10 years ago
01:35 idiosyncrat hobbs: I will admit that, when I must err, I prefer to do so on the side of overcaution.
01:37 idiosyncrat hobbs: I took the dropping of Module::Build as indicating that the Perl team itself is no longer as committed to keeping it working as it was.
01:38 idiosyncrat A central problem here is the Perl tradition of automatic upgrades.
01:39 hobbs p5p was *never* the least bit interested in keeping it working, which is part of why putting it in core was a bad idea that was reversed
01:40 hobbs but that doesn't mean that it's not maintained, it means that its maintenance is now with a more appropriate group
01:40 idiosyncrat hobbs: Right, which is why I feel a little bit blind-sided.  If I'd known there were caveats in their support, I never would have used Module::Build.
01:40 idiosyncrat their == the p5p team's
01:42 idiosyncrat Btw, I'm sorry if I'm coming across a bit negative, but it's just a number of questions that elicit references to bad experiences have come up at once.
01:42 hobbs being maintained by p5p isn't a mark of quality or reliability for a module. More the opposite.
01:44 idiosyncrat hobbs: Granted the 1st part.  (The 2nd part is further than I'd like to go.)
01:45 idiosyncrat But I *have* to rely on the p5p team if I use Perl.  Even is another team has provably better QA, having to rely on them as well *increases* risk.
01:45 idiosyncrat s/Even is/Even if/
01:46 idiosyncrat hobbs: That is, each new team that I rely on is a new point of failure.
01:47 idiosyncrat So adding dependencies on teams with better and better QA, does not decrease risk -- it just increases it more slowly.
01:47 hobbs most of what you hear about *single* points of failure isn't so great :)
01:48 idiosyncrat A lot depends on whether you're OR'ing risk or AND'ing it. :-)
02:49 ilbot3 joined #marpa
02:49 Topic for #marpa is now Start here: http://savage.net.au/Marpa.html - Pastebin: http://scsys.co.uk:8002/marpa - Jeffrey's Marpa site: http://jeffreykegler.github.io/Marpa-web-site/ - IRC log: http://irclog.perlgeek.de/marpa/today
04:46 jdurand joined #marpa
05:08 ronsavage joined #marpa
05:46 idiosyncrat Good night!
07:46 sirdancealot joined #marpa
13:00 ceridwen joined #marpa
14:32 kaare_ joined #marpa
15:43 ceridwen joined #marpa
16:31 idiosyncrat joined #marpa
16:32 idiosyncrat I now am down to 2 bugs on the Marpa::R2 list.  Those also have fixes, which are under review.
16:35 idiosyncrat I'd left several of the bugs on the list for a long time, because they were about the ASFs (Abstract Syntax Forests).  When I leave a sub-project, after a time I'm no longer able to immediately understand discussions relevant to it ...
16:35 idiosyncrat and I often wind up reading my own documentation.  Which is a good test of the docs. :-)
16:36 idiosyncrat To deal with the ASF bugs, I needed to reread my own ASF doc, and figure out what I was trying to do at the time.
17:13 yusrideb joined #marpa
19:40 sirdancealot joined #marpa
21:59 ronsavage joined #marpa
22:33 yusrideb joined #marpa

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