| drobilla | rgareus: The 'feedback' thing is because this mechanism is piggy-backing on ingen's port value stuff. It probably needs twiddling a bit for this, e.g. using ingen:activity instead, at least. | 03:05 |
|---|---|---|
| drobilla | rgareus: and I guess some kind of 'do not send update back to originating client' thing | 03:06 |
| drobilla | rgareus: The missed event, I don't know yet. Tricky with the meters, there's so much traffic. | 03:06 |
| * drobilla => bed (before 11!) | 03:10 | |
| *** thorwil has joined #ingen | 08:17 | |
| *** edogawa has joined #ingen | 09:01 | |
| *** cdr is now known as dongle | 10:15 | |
| *** edogawa has quit IRC | 11:30 | |
| *** edogawa has joined #ingen | 11:35 | |
| *** edogawa has quit IRC | 11:40 | |
| *** edogawa has joined #ingen | 11:45 | |
| *** thorwil has quit IRC | 14:03 | |
| *** thorwil has joined #ingen | 16:42 | |
| Anchakor | drobilla: is the new turtle really so much pain? I've been thinking it would be good to have w3c independent set-in-stone human friendly serialization, so if you decided to just give up on turtle and make it, it could be good | 17:53 |
| rgareus | drobilla: how do I run an ingen-patch in jalv? | 17:55 |
| drobilla | Anchakor: Not if you're just copying rules into a lax parser generator, I suppose. | 17:56 |
| rgareus | drobilla: I've created an ingen patch, saved it to ~/.lv2 ; lv2ls shows it, but jalv.gtk gives me lilv_plugin_get_library_uri(): warning: Plugin <file://....> has no lv2:binary | 17:57 |
| drobilla | Anchakor: If you're implementing a recursive descent parser by hand, it has increased readahead requirements, and a massive amount of hairy special cases. It's much worse to implement. | 17:57 |
| drobilla | rgareus: You need the latest lilv, to support lv2:prototype. I removed the symlink thing. | 17:57 |
| drobilla | Anchakor: I'm strongly inclined to make one, but need Turtle compatibility anyway, so it's just more work. | 17:58 |
| drobilla | Anchakor: Maybe if I ever decide to go off on some epic wank of an N3 rules tangent | 17:59 |
| drobilla | I'm also pretty inclined to make an s-expression one, because, well, y'know... :) But a simpler superset of Turtle with trivial parsing rules would be nice. | 17:59 |
| rgareus | drobilla: mmh I thought I have latest lilv. and right it does not work in ardour, either. | 18:00 |
| Anchakor | well you could make the turtle parser using lex & yacc or whatever and people who wanted a serialization with efficient streaming parser/serializer they would use the one you make | 18:01 |
| drobilla | rgareus: Hm, you need the ingen bundle installed, is that it? | 18:01 |
| Anchakor | I'm more inclined to s-expression syntax too :) | 18:02 |
| drobilla | Suits me :) | 18:03 |
| drobilla | I kind of want to revive the n3 spirit. Sometimes I fantasize about giving it a name, buying a domain, having simple tutorials. | 18:03 |
| drobilla | Ditch all the ugly XML shit, all the bureacracy, all the SQL shit, all the enterprisey shit in general. Nice, simple, understandable. | 18:04 |
| Anchakor | I'm not really into extending the core model as n3 does | 18:05 |
| drobilla | Well, not that part. | 18:05 |
| Anchakor | I'd rather keep it as minimal as possible to build the rest over it | 18:05 |
| drobilla | Definitely entirely data model based, I agree. | 18:06 |
| drobilla | and homoiconic | 18:06 |
| drobilla | No fucking SQL because some logical theory atronomics says there's semantic problems with quads or whatever | 18:07 |
| drobilla | (That said I personally have no need for queries anyway) | 18:07 |
| Anchakor | well rdf is homoiconic sort of by definition of being a data language | 18:08 |
| rgareus | drobilla: where does lilv look for the ingen bundle? | 18:08 |
| Anchakor | so homiconic are all its serializations | 18:08 |
| drobilla | Yes, but SPARQL is not. | 18:08 |
| Anchakor | yeah, but sparql is crap | 18:08 |
| drobilla | Anchakor: indeed. Yet, it drives the whole thing, these days. The reason Turlte is not a fucked up ugly mess is "SPARQL did it therefore fuck you" | 18:09 |
| drobilla | rgareus: Anywhere in LV2_PATH, the usual | 18:09 |
| drobilla | Anchakor: The worst by far is that the grammar depends on SPARQL path queries. They mashed the two tokens on top of each other with no distinguishing thing. | 18:09 |
| rgareus | drobilla: it is in LV2_PATH | 18:09 |
| drobilla | Anchakor: So now we have a data syntax with this big list of arbitrarily disallowed characters in names because they're part of a query syntax that isn't in the language? WTF? | 18:10 |
| drobilla | How is that not *objectively* incompetent by any sane measure? | 18:10 |
| rgareus | drobilla: under ~/local/lib/lv2/ingen.lv2 LV2_PATH includes $HOME/local/lib/lv2/ | 18:10 |
| * rgareus BBL | 18:11 | |
| drobilla | rgareus: Not sure, it should be finding the prototype then | 18:11 |
| Anchakor | drobilla: it's pretty ridiculous, like if people made json2 and forbidden some characters in the property names because some widely used REST API used it in a special way | 18:12 |
| drobilla | Anchakor: Yep, totally absurd. The SPARQL WG fucked up. | 18:12 |
| drobilla | Anchakor: Then the overall management of RDF standardization fucked up by standardizing SPARQL first, and letting those fuckups that didn't consider any of these problems ruin Turtle | 18:13 |
| Anchakor | I think the problem is that the big data and OLAP people are driving the thing and it's big there | 18:13 |
| drobilla | Pretty much. Enterprisey people who could not possibly give less of a shit about elegance. | 18:13 |
| Anchakor | but w3c is full of problems like this with little to no consistency between work of different WGs | 18:14 |
| drobilla | (and clearly have no business designing grammars in the first place) | 18:14 |
| drobilla | It would be a nice jab to make a sane syntax and have no reservations about explaining why... | 18:14 |
| Anchakor | though I dunno if s-expression syntax wouln't look too outlandish. people are used to infix operators | 18:16 |
| drobilla | Yes, that would be pushing it. | 18:17 |
| drobilla | The Turtle thing does have the problem that it actually incentivises using blank nodes, though. | 18:17 |
| drobilla | Because you can abbreviate those inline, but you can't resources with URIs. | 18:17 |
| drobilla | Which is exactly why LV2 ports are almost always blank nodes, as it happens. | 18:18 |
| Anchakor | well what about forbidding unnamed "bnodes" and then require a bnode base uri which they would use as prefix uri; real blank nodes would be some ridiculous thing, like '<blank-node>' :) | 18:20 |
| drobilla | The semantics astronauts would say that's abuse of URIs because they aren't really global URIs. | 18:23 |
| Anchakor | well they would be, with the bnode base uri | 18:24 |
| drobilla | But either way, that doesn't avoid the problem. | 18:24 |
| drobilla | The goal is to be able to inline things and give them an explicit URI. | 18:24 |
| drobilla | A consistent syntax for a resource description, basically. | 18:24 |
| drobilla | Maybe graftable onto a Turtle-like syntax with a keyword after [ | 18:24 |
| Anchakor | I don't think it's possible without naming the resources | 18:24 |
| drobilla | A better syntax would just use the same syntax for describing resources everywhere. | 18:25 |
| Anchakor | else the reorder your statements and the uris are changed | 18:25 |
| drobilla | {id :name "David" :awesomeness 9000} or whatever. | 18:25 |
| drobilla | where id can be [] or whatever if there isn't one | 18:26 |
| Anchakor | yeah with my idea that would be like bnodeprefix:id | 18:27 |
| Anchakor | but the id has to be there and the user must not change it over time | 18:27 |
| drobilla | Orthogonal, but not a bad idea, if you want to literally eliminate blank nodes. | 18:27 |
| drobilla | Though the order thing like you said means it doesn't really achieve that. | 18:28 |
| drobilla | Well, it kinda does, I guess, they just don't have stable URIs | 18:29 |
| Anchakor | well with [] you have problems with order. with _:bnodelabel not | 18:29 |
| drobilla | Right. My gripe is with the syntax though. | 18:29 |
| drobilla | If you want to inline it, you don't get to give a name of either variety. This is shit. | 18:29 |
| Anchakor | well I thought you were dismayed people used bnode for plugin ports | 18:30 |
| drobilla | Well, a bit. Maybe with LV2 for example we should have just made all the ports top level descriptions | 18:30 |
| drobilla | Yes, but the reason they do is that you can inline them so it makes a more readable/terse definition, and makes the whole plugin definition one structure | 18:30 |
| drobilla | i.e. Turtle gives you incentive to use blank nodes | 18:30 |
| drobilla | We even mandate a property to give a symbol, which could be used to construct a valid URI anyway! | 18:31 |
| drobilla | It's totally just a failure of the syntax | 18:31 |
| Anchakor | well a syntax which made using proper uris (using prefixes) as friendly as possible and discouraged using blank nodes would be good - do we agree? | 18:32 |
| Anchakor | my idea is that then people would make garbage uris if they were forced to make uris so I think it would be useful to separate the garbage uri space with the proper uri space - thus special prefix for pseudo-bnodes | 18:34 |
| dongle | i'd be happy if every library came up w/ identical bnode identifiers via some std git-style hashing | 19:58 |
| dongle | the _:b27374 thing is annoying as it instantly throws 'bad prefix' errors everywhere | 19:59 |
| dongle | maybe thats the idea so you dont accidentally serialize them but ugh anyway | 19:59 |
| drobilla | Sounds more like a problem with toolchain and/or workflow than anything | 20:03 |
| drobilla | I mean, _:bnnnn pretty much is a/the standard bnode style | 20:04 |
| Anchakor | dongle: that's not possible, the hashing thing, bnodes can be identified only by their relationships which change, and even then its not unique identification | 20:06 |
| dongle | order and sha256 the turtle of their contents? | 20:10 |
| dongle | filesystems hve forced ppl to come up w/ names since..forever | 20:11 |
| dongle | not really interested in inventing another rdf serialization or toolkit | 20:11 |
| drobilla | Ugh, hashing. | 20:11 |
| drobilla | Give it a sensible identifier or just deal with the fact that it doesn't have one. | 20:12 |
| drobilla | Though it doesn't matter much in this case since it wouldn't work anyway. | 20:12 |
| *** thorwil has quit IRC | 21:12 | |
| *** edogawa has quit IRC | 23:28 | |
Generated by irclog2html.py 2.13.0 by Marius Gedminas - find it at mg.pov.lt!