This is a public chat log generated from the #semsol IRC channel.
23:47:13
nn
00:10:21
i removed the @ as bengee suggested
00:10:30
got error message matching this issue: http://www.somacon.com/p109.php
00:12:07
Looks like php-mysql got lost somewhere, when i removed another php package
00:12:09
=============================================================================
00:12:10
Package Arch Version Repository Size
00:12:11
=============================================================================
00:12:12
Installing:
00:12:13
php-mysql i386 5.0.4-10.5 updates-released 60 k
00:12:15
Installing for dependencies:
00:12:16
gmp i386 4.1.4-6 base 593 k
00:12:17
php i386 5.0.4-10.5 updates-released 2.3 M
00:12:18
php-pear i386 5.0.4-10.5 updates-released 371 k
00:12:19
Transaction Summary
00:12:20
=============================================================================
00:12:22
and an apache restart seems to have fixed it!
00:12:50
so, my doh there... (though more guidance from error messages would be sweet)
10:53:31
danbri, glad you managed to hunt the issue down, I'll remove the @ in the next rev for better error reporting
12:00:04
cheers!
15:07:04
danbri: I actually had the same issue with arc1 a few months ago!
15:07:29
php mysql problem without error reporting?
15:07:32
I reported to bengee who said he do a better error reporting feature for arc2 :p
15:07:37
danbri: yes
15:07:54
and so he did
15:08:08
it was the same thing: no error anywhere - but nothing was working
15:08:09
today :)
15:08:13
forgot that @, though
15:08:19
:D
15:08:54
danbri: I guess you didn't have any other php-mysql script running on the same machine, did you?
15:09:04
such as phpmyadmin or drupal...
15:10:30
no, its a clean sandbox i just set up
15:10:46
well, its dirty in that i am backing up only the scripts
15:10:47
amazon ec2
15:11:02
i should really burn myself an AMI image with the core config established
15:11:16
http://sandbox.foaf-project.org/2008/foaf/ggg.php
15:18:18
bengee off for dance lessons, back in 2 hours
15:23:36
kwijibo, I'm due to do a podcast with bengee soon (maybe this week)
15:24:25
iand just suggested I ask you for good questions to ask him, re. arc etc
15:24:46
danja, sounds really cool
15:25:27
I have a question
15:25:28
I'd ask him stuff on why PHP / MySQL for RDF Web development
15:25:35
why RDF for web dev
15:25:45
sorry danbri - on you go
15:26:00
"If ARC bugs are so easy to fix (like they're fixed in 2 mins of me reporting them usually) why do they exist in the 1st place!"
15:26:29
seriously, the fix cycle on arc is v impressive
15:26:31
I'd guess... no unit tests?
15:26:42
i'm just being cheeky, it's great...
15:26:50
:)
15:27:04
it is
15:28:03
danja: get him to explain about why RDF/SPARQL helps RAD
15:28:20
emm, target audience for ARC
15:28:28
applications built on arc
15:29:00
the plugin architecture for arc
15:29:20
all about Trice
15:29:47
I dunno, loads of stuff :)
15:31:17
also, in parallel to why rdf for webdev, why NOT rdf for webdev? when isn't it the right choice, and why does he think the uptake hasn't been faster I guess
15:32:01
views on uppercase vs lower-case semantic web
15:32:17
semantic html
15:33:05
when to use µf vs GRDDL vs RDFa vs eRDF vs custom formats
15:37:32
also, the obvious one about ARC's aim to make RDF dev accessible for shared hosting LAMP stacks with small open source PHP/MySQL stores, vs Talis aims to make it easier with big scalable REST/RDF platform services
15:39:23
because it's kind of like two approaches to the same problem
15:39:30
or could be viewed that way
15:39:55
so are the two in competition somehow?
15:42:31
danja: if you ask him half of that, it'd be a pretty interesting podcast from my POV ;)
15:47:56
...and mine, thanks!
17:44:55
so bengee, what questions would you ask you, if you were danja? ;)
17:53:20
¡Hola!
17:54:37
bengee: I checked with Niklas Lindström re normalizing element content when parsing RDFa.
18:17:07
heh, cool scrollback
18:17:18
peterkz, and Niklas is right, I assume?
18:18:15
I would think so. He is looking for the reference (since I couldn't find it myself)...
18:18:30
I'll let you know as soon as I get it.
18:18:46
I can prolly make the literal check more tolerant
18:19:16
ad only use an xmlliteral datatype when there are tags in the value
18:19:22
s/ad/and/
18:20:36
kwijibo, those are great questions
18:21:16
now we need someone else to provide answers, and danja and me are going to have a great time during the podcast ;)
18:21:33
lol
18:25:16
That sounds great.
18:28:50
bengee, is there any way to trigger a construct query within arc2 whenever a sparql update is done?
18:29:15
could be inefficient i know
18:29:22
peterkz, it's probably less a normalization issue but rather an xml detection one, and looking at the rdfa syntax doc, I should not set the xml literal flag when newlines occur
18:29:23
am thinking to do groups stuff that way
18:30:27
"the current object literal is a plain literal if the element contains only text"
18:30:35
bengee: Ah, you found it. Great!
18:30:37
which it does in your case, I think
18:30:54
Yes.
18:30:55
I'll fix that
18:31:27
Great! Sorry for my poor bug reporting skills...
18:31:33
nah
18:32:27
danbri, you mean that an update returns the result of a related CONSTRUCT?
18:32:52
or that the construct result is stored?
18:33:23
sorta like "something changed, update this graph"
18:35:34
bengee: I now have a working script that will scan a folder (recursively) and load all files (of certain types) into their own named graphs (based on the file path) while adding the modification time to a related named graph.
18:35:51
awesome!
18:36:07
This may be useful for people who keep data in the filesystem and want them synced with the triple store.
18:36:21
ooh, that sounds sweet
18:36:41
yeah, I chatted a bit with kwijibo about scenarios for that
18:36:55
I will try to move the code into an ARC plugin (right now it isn't very pretty - I am new to PHP).
18:37:25
I can help you with the plugin, if necessary
18:37:43
would make a great one
18:38:29
imagine a rather large triple store, but you'd like to have fast response times fro editing, and asynchronous index updates
18:39:40
you could "check out" a graph to a local file, edit it w/o index update problems, save it, and then have it auto-loaded back into the public rdf store
18:41:01
that could really solve the editing problem resulting from query-optimized large indexes
18:41:06
bengee, i mean that the server does some private housekeeping whenever it sees updates
18:41:14
...needn't return anything for that interaction
18:41:23
'k, I see
18:41:27
eg. if i update my xfn-export from homepage named graph
18:41:47
... the server recomputes a private named graph that generates foaf:Group of crushes and so on
18:42:04
and then I envisage user-facing apps being built against a private to owner set of foaf Groups
18:42:23
'cos nobody really manages their twitter and flickr buddylists in a way that corresponds to how they wanna filter their views
18:43:01
peterkz - folder scan thing sounds cool
18:43:09
would be nice to have the named graph more configurable tho
18:43:29
was chatting with ldodds earlier about wiring up his Slug foaf crawler to arc and other triple stores, and would need something like that
18:43:34
bengee - am i making any sense? :)
18:44:13
danbri: If you can name it from something in the file that could work to I guess.
18:44:53
Right now the graph name is created by a simple method that just concats "file:" with the path.
18:45:07
The principle is fetched from OORT: http://oort.to/oort_api/oort.util.graphs-pysrc.html
18:49:14
danbri, I guess should add a new "triggers" directory to ARC then
18:49:41
if you think the idea makes sense
18:49:46
I like it anyway :)
18:49:52
is similar to my 4am blog post last night
18:49:54
or related
18:50:11
a crawler would unpack some queries it finds in the web (from the right people anyway) and store as triggers
18:50:32
so groups could be defined in terms of workplaceHomepage or xfn:child or whatever
18:51:00
it might be a kinda nerdy feature without some UI but maybe a simple authoing tool is possible? not sure
18:52:55
curent plan is to keep ARC core as low-level as possible
18:53:30
Trice will have store management, access control, forms and everything
18:53:52
but a drop-in folder for triggers should be doable
18:54:33
plus a config setting/flag, so that ARC doesn't have to check for triggers on each request even when none are defined
18:56:14
so, you could say $config = array(....., insert_triggers => array('rebuildWhitelist', 'sendEmail'))
18:57:24
and then there'd be a ARC2_rebuildWhitelist.php in the triggers dir
18:57:35
something like that
19:07:30
is trice opensource too?
19:07:46
yes
19:07:48
drop in folder for triggers, cool :)
19:08:13
danbri heads out for evening ... bye for now
19:08:16
happy hacking!
19:08:34
bengee updates http://arc.semsol.org/dev with RDFa bug, and triggers request
19:10:01
happy eveninging :)
19:27:11
Sorry for the delay (diaper change). The raw hack for folder sync is here: http://pastie.caboo.se/142002
19:27:48
Oh. Diaper change was for my son obviously, if you thought otherwise...
19:27:54
:-)
19:30:29
heh
19:31:38
script looks great
19:32:44
It needs more testing and cleanup I am sure (as I said I am struggling with PHP).
19:33:58
But I ran it for a couple of hundred docs here and it worked for me.
19:34:25
cool
19:35:29
there are getSetting and setSetting methods in ARC, in case you don't need the modification time in other situations than the script
19:35:35
Only thing I wonder is limits for long running PHP processes. Most shared hosts I know of have had a time limit.
19:35:46
Ah cool.
19:35:52
yeah, usually a couple of minutes CPU time
19:36:26
one thing you should add is a set_time_limit()
19:36:31
s/should/could/
19:36:53
this will reset the default timeout of 60(?) secs
19:37:51
*looks into PHP docs*
19:37:54
set_time_limit(0) => no limit
19:38:27
Ah, that should do the trick.
19:38:49
then you are at least provider-bound, not config-bound
19:39:38
Yeah, SPARQL query time will be significant also. It shoud be possible to store the modification time in a plain sql table as well (for speed) but it wouldn't be very elegant:-)
19:43:15
Gotta go. See ya!
19:43:27
cya
19:49:48
bengee thinks his rdf editor is finally working
19:50:35
boy, that smushing stuff is tricky
19:50:41
yep :)
19:52:23
I *think* ARC can now handle bnode editing, even when the bnode in question has been "smushed away"
19:52:44
it works in the editor, for some magical reason
20:01:06
mortenf tries to use group by in sparql
20:01:29
bengee ducks
20:02:03
it doesn't always translate to sql so easily
20:02:12
a DISTINCT helps sometimes
20:02:20
not here
20:02:33
seems to work :)
20:02:44
hey
20:07:57
bengee wonders where to put his foaf file/wiki thingy
20:08:30
hmm, bnode.org/data/bengee#self
20:09:31
or /stuff/.. or /wiki/...
20:31:41
bengee is now http://bnode.org/grawiki/bengee#self and calls it a day
21:14:10
defin'ly something wrong with the smusher.
21:14:43
test case: load our two foaf files, and the foaf schema, smush.