06:18:55 <jonkri> stepcut: Do you have any plans to convert the XML types of HSP to those of xml-types?
12:24:18 <donri> sm: btw your feed on the planet is a bit messed up, no formatting not even for paragraphs
12:24:56 <donri> so every post shows up in my reader as one large paragraph of plain text
12:25:37 <donri> hm looks good on the planet web site though
12:30:08 <donri> sm: looks like you're using <summary type="html"/> but from the spec it looks like summary is always text and you want the <content> element with that attribute
14:33:05 <sm> donri: thanks for letting me know. Are you sure ? http://feedvalidator.org/check.cgi?url=http%3A%2F%2Fjoyful.com%2Fblog.xml validates it, and it looks fine in my reader (bazqux)
14:34:14 <donri> sm: https://tools.ietf.org/html/rfc4287#section-4.2.13
14:34:35 <donri> compare https://tools.ietf.org/html/rfc4287#section-4.1.3
14:36:33 <sm> thanks, testing
14:36:34 <donri> sm: actually you're right https://tools.ietf.org/html/rfc4287#section-3.1
14:36:53 <donri> sm: still, summary should be a summary not the complete content ;)
14:41:05 <sm> yeah I think a html summary is legal. I agree that content seems more appropriate, unfortunately this requires changing hakyll
14:41:14 <sm> what's your reader ?
14:41:30 <donri> sm: currently https://theoldreader.com
14:44:34 <sm> http://joyful.com/blog2.xml uses content, if you can test.. looks good here
14:45:03 <donri> sm: works great
14:50:42 <sm> donri: hakyll's renderAtom doesn't really know if you're giving it the full item content or not, so it may not be appropriate to just use content. We should probably bring it up on #hakyll
14:51:17 <sm> you could also get theoldreader to handle it better (it's legal after all)
14:52:11 <sm> or switch to bazqux! haskell-based!
14:52:13 <sm> just kidding
14:55:21 <donri> sm: quite, it's a bug in theoldreader
14:55:31 <donri> sm: might switch to bazqux once they add autorefresh
14:55:52 <donri> kinda want something open source though, but haven't found anything to my satisfaction
14:56:17 <sm> I hear you. The maintainer did say on hacker news he'll open the source at some point
14:57:59 <donri> yeah, but that's sort of missing half the point ;)
14:58:17 <sm> how so ?
14:59:09 <donri> well half the point is "the community project", ie. what open source is about
14:59:23 <donri> the other point is about user rights, ie. what free software is about
15:00:28 <sm> hopefully he'll make it free software. I forget what he said exactly
15:02:02 <sm> I feel I should pay down my off-topic debt a little, here goes..
15:03:45 <sm> darcsden has HSP templates all in a separate directory. I think there's no advantage to this, it'd be easier to have them together with or inside the snap handlers. Any opinions from other projects ?
15:07:30 <donri> sm: just treat the "templates" like functions and consider how you'd structure that normally :)
15:07:52 <donri> ie. you might abstract common parts into functions but write simple stuff inline
15:08:06 <sm> donri: exactly
15:08:12 <donri> ie. you might move collections of functions into dedicated modules or just put them in the module that uses them
15:09:28 <sm> this MVC-ish file layout is common, but it just means I have to juggle twice as many files and fiddle with template parameters in two places
15:10:18 <donri> yeah. complete separation makes more sense if you have some logic-free template engine like heist
15:10:57 <sm> ..which I'm thinking of trying, for the runtime reloading
15:11:27 <donri> you could use plugins{,-auto} to runtime [re]load hsp
15:11:42 <donri> it's not awesome though because it doesn't use cabal
15:12:53 <donri> or maybe web-plugins, not sure how it works
15:13:22 <sm> it those are reliable and have an easy example to follow, I might try it
15:13:28 <sm> if
15:14:11 <donri> i think stepcut wrote web-plugins for clckwrks which probably uses it
15:14:31 <donri> but it's not tied to anything, all platform deps
15:14:59 <donri> http://hub.darcs.net/stepcut/web-plugins/browse/example
15:16:17 <sm> one of the things that's great about HSP is how it automatically picks up bindings in scope
15:17:20 <sm> (vs. heist). I guess the shakespearean templates also do that.
15:17:24 <donri> stepcut: does web-plugins do runtime recompilation?
15:17:35 <donri> sm: yeah. and blaze-html
15:17:40 <sm> yeah
15:18:20 <sm> but see, I don't want recompilation - I already have that (using searchpath). Being able to tweak ui without the linking delay is what's attractive about heist
15:18:24 <donri> obviously there are benefits to the heist approach as well, particularly for non-programmers
15:19:07 <donri> sm: the plugins package works via dynamic loading though, not sure there'll be a big linking delay
15:20:09 <sm> ok, I probably misunderstand how that works in practice
15:20:10 <donri> still has to compile the code though, although there's ways to speed that up too of course
15:46:51 <stepcut> donri: there is a web-plugins-dynamic package which has proof-of-concept dynamic reloading, but is not at all integrated with cabal for file watching
15:47:07 <stepcut> donri: some day it will though :)
15:47:45 <stepcut> donri: it's not the highest priority clckwrks issue at the moment
15:48:24 <stepcut> sm: this is what you can do with plugins, http://www.youtube.com/watch?v=ok_gf0QVLnE
15:48:37 <stepcut> sm: except it has bitrotted I am sure
16:01:27 <donri> sm: btw hub uses snap right, doesn't it have an interpreted mode you could use?
16:02:06 <donri> otherwise, run it from ghci and :r ;)
16:07:43 <sm> donri: that seems.. unlikely to work but I will try :)
16:09:02 <donri> sm: if you mean ghci you'd have to use :main ^C :reload :main so it's some steps to type but the loading might be faster than a full build
16:11:36 <sm> ghci == the interpreted mode you were talking about ?
16:12:02 <donri> sm: well not quite; i think snap has a special interpreted mode via the hint package
16:12:15 <donri> in the end it works like ghci (ghc interpreter) but automated
16:12:56 <donri> sm: see bottom of http://snapframework.com/docs/quickstart
16:13:03 <sm> oh, thanks for the tip
16:13:05 <donri> so some steps might be necessary to make it work with hub
16:14:08 <sm> yes hub/darcsden are snap- and HSP-based .
16:14:27 <donri> well i think this is a feature of the snap *server*
16:14:46 <donri> but better ask #snapframework
16:14:50 <donri> i haven't used it
16:15:25 <donri> sm: http://hackage.haskell.org/package/snap-loader-dynamic