2004-07-21T08:41:03Z -!- AndyS [AndyS@192.6.10.2] has joined #dawg 2004-07-21T13:43:47Z -!- DaveB [dajobe@137.222.34.57] has joined #dawg 2004-07-21T14:15:54Z -!- DanC [connolly@18.29.0.30] has left #dawg [Client exiting] 2004-07-21T14:30:55Z -!- ericP changed the topic of #dawg to: DAWG test suite and documents -- 21 July 2004-07-21T14:34:50Z andy, you 'round? 2004-07-21T14:36:51Z -!- SteveH [swh@152.78.190.249] has joined #dawg 2004-07-21T14:36:59Z hi steve 2004-07-21T14:37:05Z hi, sorry I'm late 2004-07-21T14:37:12Z np 2004-07-21T14:37:19Z do you have a w3 web account? 2004-07-21T14:37:25Z I dontthink so 2004-07-21T14:37:44Z * ericP looks in the db... 2004-07-21T14:37:58Z my w3c username is sharris2 2004-07-21T14:38:08Z oh, that's what i meant 2004-07-21T14:38:17Z ah, ok 2004-07-21T14:38:18Z are you content with that for your CVS name? 2004-07-21T14:38:31Z I would prefer "swh", but its no big deal 2004-07-21T14:38:39Z ok 2004-07-21T14:39:47Z do you know how you got this account? 2004-07-21T14:39:58Z did you apply for member access or somehting like that? 2004-07-21T14:40:29Z I think it was created as a side effect of joining DAWG 2004-07-21T14:40:37Z the sharris2 account has no details so i'm trying to see what script needs fixing 2004-07-21T14:40:45Z DanC may have made it 2004-07-21T14:40:51Z (the acount) 2004-07-21T14:41:33Z i'll follow up with him -- tx 2004-07-21T14:42:51Z after "swh", what's your 2nd choice? (filling out a form with one alternate choice) 2004-07-21T14:43:06Z sharris2 I guess :) 2004-07-21T14:43:24Z or steveh 2004-07-21T14:43:58Z Ping - sorry I'm late 2004-07-21T14:44:07Z hi AndyS 2004-07-21T14:44:36Z howdy andy 2004-07-21T14:44:51Z Eric - I have a ssh key question later. 2004-07-21T14:45:09Z steve, should we use S.W.Harris@ecs.soton.ac.uk for the followup email? 2004-07-21T14:45:12Z andy, roger 2004-07-21T14:45:35Z So - testing - what do should we cover today? 2004-07-21T14:45:42Z ericP: yes please 2004-07-21T14:45:55Z Sorry - I'll let you finish. 2004-07-21T14:46:20Z np 2004-07-21T14:49:40Z steve, got a snailmail address i can associated with this account? (/msg me with it fi you want) 2004-07-21T14:49:59Z User identification info (Address of account user, phone and company information, etc.): 2004-07-21T14:50:09Z is the input i'm trying to fill in 2004-07-21T14:50:29Z ericP, Zepler Building, Univerity of Southampton, SO17 1BJ, UK 2004-07-21T14:51:22Z phone? 2004-07-21T14:51:26Z ericP, should all be here: http://www.ecs.soton.ac.uk/people/swh 2004-07-21T14:52:01Z tx 2004-07-21T14:52:27Z nb my Resarch Staff phone number is wrong 2004-07-21T14:53:30Z i gave xxxxxxxx.2831 2004-07-21T14:53:45Z its .8347 2004-07-21T14:53:59Z * SteveH should get that fixed 2004-07-21T14:55:20Z updated 2004-07-21T14:55:30Z ok, we probably have a couple days wait on that 2004-07-21T14:55:33Z thx 2004-07-21T14:55:45Z steve, you are CC'd on the request (well, the reply to it, anyways) 2004-07-21T14:56:10Z I hacked my querytest code into an HTML generator last night 2004-07-21T14:56:24Z wasnt difficult, but it doesnt spit out very good HTML yet 2004-07-21T14:56:34Z Style sheets! 2004-07-21T14:56:41Z naturally 2004-07-21T14:57:00Z by "not very good" I mean its not valid - missing header and footer 2004-07-21T14:57:10Z danc, asked me to stick those tests in the botton of the BRQL^H^H^H^Hrq23 document 2004-07-21T14:57:29Z So what does it generate? Is it running the tests? 2004-07-21T14:57:32Z ... so we'd all know exactly what we were voting on 2004-07-21T14:57:56Z AndyS, it doesnt run them, just turns them into something viewable - I took the actual testing code out 2004-07-21T14:57:56Z now that we've selected the strawman, he wants the tests back out. 2004-07-21T14:58:06Z Currently (and it can change) I have a manifest file which lists query/dataset/result expected 2004-07-21T14:58:06Z seems reasonable 2004-07-21T14:58:17Z that reduces the burden on the rq23 generation to include automatically generated HTML 2004-07-21T14:58:24Z I'l recode the old tests in Andys manifest format 2004-07-21T14:58:25Z so your html can be standalone 2004-07-21T14:58:38Z then JUnit runs each test and checks the outcome (RDF graphs match) 2004-07-21T14:58:46Z Is that compatible? 2004-07-21T14:58:47Z AndyS, the manifest format looks fine to me (though I'm not a fan of RDF lists) 2004-07-21T14:59:09Z Nor am I but I like by JUnit tests to run in the same order each time! 2004-07-21T14:59:19Z Heh 2004-07-21T14:59:21Z s/by/my/ 2004-07-21T14:59:23Z is it worth having collections? 2004-07-21T14:59:38Z doesnt hurth that much 2004-07-21T14:59:40Z Collections are worse IMHO. Look terrible in N3. 2004-07-21T14:59:41Z *hurt 2004-07-21T15:00:18Z The vocab of the manifest is two level: a "test" is an action and result 2004-07-21T15:00:25Z TestSet1 tests (_:test1, _:test2, _:test3) 2004-07-21T15:00:27Z oh, I'm parsing it with DaveB's Turtle parser, so it doesnt like float constants, but otherwise seems OK 2004-07-21T15:00:46Z yeah, Turtle doesn't have them yet 2004-07-21T15:00:48Z and a query action is a data file and query file (this was at Alberto's request I think - too long ago) 2004-07-21T15:01:28Z Jena's "Turtle" does :-) 2004-07-21T15:01:33Z i tend to use collections over bags 'cause i think they are prettier in n3 2004-07-21T15:01:59Z though the cons structure much uglier in the graph, of course 2004-07-21T15:02:12Z Could give a number as a property and look in order but a list is what it is naturally. 2004-07-21T15:02:19Z yes, parsing it with BRQL/RDQL is a pain 2004-07-21T15:02:33Z We have support code to mask such yukkies. Used a lot in OntAPI. 2004-07-21T15:02:46Z sounds like a use case for a collection accessor fucntion 2004-07-21T15:02:47Z * AndyS smiles at list in query matters 2004-07-21T15:03:12Z and the match to SQL is - err - not nice 2004-07-21T15:03:23Z true 2004-07-21T15:04:16Z steve, can you give us an idea of your file tree? 2004-07-21T15:04:20Z Containers would be as bad. Could (shock, horror) use URIs with a number but lists are the least worst. 2004-07-21T15:04:32Z perhaps zip and send it to us? 2004-07-21T15:04:37Z ericP, file tree? 2004-07-21T15:04:53Z you mean the script that generates the .html? 2004-07-21T15:05:10Z yeah, your lists of tests, a manifest, script, whatever you're running now or want to move to 2004-07-21T15:05:47Z ok, copying it to somewhere public 2004-07-21T15:05:58Z i have an input file (all the rq23 tests run on the same data), a set of input files and a set of output files 2004-07-21T15:06:03Z all in the same directory 2004-07-21T15:06:16Z no makefile, no script 2004-07-21T15:07:26Z http://www.ecs.soton.ac.uk/~swh/TestSketchCases/ has the product, i assume 2004-07-21T15:07:30Z http://www.ecs.soton.ac.uk/~swh/GenTestCases/ 2004-07-21T15:07:33Z or something close 2004-07-21T15:07:34Z tx 2004-07-21T15:07:36Z foo.html is an exaple output 2004-07-21T15:07:52Z still a bit rough round the edges 2004-07-21T15:08:00Z but Il neaten it up this evening 2004-07-21T15:08:30Z That's pretty neat 2004-07-21T15:09:48Z Alberto will like the Perl :-) 2004-07-21T15:10:19Z yeah *cough* I havea bit of a perl habit thats hard to kick 2004-07-21T15:10:37Z i really love perl for portability and debugability 2004-07-21T15:10:51Z java debugger can't compete at this point 2004-07-21T15:11:05Z i'm getting my file list together to show you guys... 2004-07-21T15:11:41Z How do you debug a perl script? Eclipse debugger is point-and-breakpoint 2004-07-21T15:12:12Z nothing beats valgrind + gdb for debugging :) 2004-07-21T15:13:03Z "grind" is the word. 2004-07-21T15:14:33Z valgrind's great but not cross-language or with threads 2004-07-21T15:15:24Z the beta handles threads much better 2004-07-21T15:15:28Z re debugging in perl, i use perldb mode in emacs 2004-07-21T15:15:46Z but it sounds like i have to check out Eclipse for java 2004-07-21T15:15:55Z here's my file tree: 2004-07-21T15:15:56Z defintely 2004-07-21T15:15:59Z -!- Irssi: Pasting 8 lines to #dawg. Press Ctrl-K if you wish to do this or Ctrl-C to cancel. 2004-07-21T15:16:01Z in WWW/2001/sw/DataAccess/rq23, i have (mostly not checked in 'cause i don't know the final file names): 2004-07-21T15:16:01Z Overview.html -- rq23 spec 2004-07-21T15:16:01Z prod2Table -- script to generate the tables of grammar productions 2004-07-21T15:16:01Z dawg-data-01.n3 dawg-data-01.rdf -- data for the following tests 2004-07-21T15:16:01Z ex_1.brql ex_2.brql ex_3.brql ex_4.brql -- examples from the rq23 document 2004-07-21T15:16:01Z dawg-query-001 dawg-query-002 dawg-query-003 2004-07-21T15:16:01Z dawg-result-001.n3 dawg-result-002.n3 dawg-result-003.n3 2004-07-21T15:16:01Z dawg-tests.n3 -- manifest - i'm not using it right now 2004-07-21T15:16:04Z . 2004-07-21T15:16:24Z well, not much of a tree so much as a field of grass 2004-07-21T15:17:11Z are the dawg-*.n3 files the same ones I have from Andy? 2004-07-21T15:17:43Z yup 2004-07-21T15:18:13Z No tin CVS at the moment as far as I can see - just did "cvs update" 2004-07-21T15:18:25Z of those names, i have invented Overview.html, prod2Table, dawg-data-01.rdf, ex_{1,2,3,4}.brql 2004-07-21T15:18:46Z yeah, waiting to figure out what to call them 2004-07-21T15:19:07Z seems like *.brql has to change anyways 2004-07-21T15:19:10Z Woudl be good to split examples from tests - maybe now, maybe later 2004-07-21T15:19:42Z wouldn't examples belong in the rq23 doc? 2004-07-21T15:19:51Z *.rq? c.f. *.xq (which I see for XQuery's) Any known use of .rq? 2004-07-21T15:20:12Z Examples - yes - but we can generate the HTML fragments. 2004-07-21T15:21:28Z there's a file that correlates file extensions to mime types defined in /etc/mailcap. anyone know the name? 2004-07-21T15:21:29Z Or maybe these are tests as well? 2004-07-21T15:21:38Z .rq is not in magic 2004-07-21T15:21:50Z ericP, /usr/share/file/magic 2004-07-21T15:22:24Z Isn't that magic numbers to file types? 2004-07-21T15:22:34Z .rq isn't defined in my Windows registry. 2004-07-21T15:22:46Z AndyS, its both on my system 2004-07-21T15:25:39Z Logs into Linux/RH - types "less magic" ... 2004-07-21T15:25:40Z * ericP gives up looking for ".mpg" in /etc/{mime-,}magic -- find /usr/share/misc/file/magic not trivial to understand 2004-07-21T15:26:27Z steve, looks like our two day wait is done 2004-07-21T15:26:35Z neat 2004-07-21T15:27:01Z those seconds just dragged on into minutes 2004-07-21T15:27:36Z anyone have a proposal for a file tree or web tree? 2004-07-21T15:28:17Z from a web perspective, we can randomly name everything and put it anywhere we want, but that's sort of a burden on Makefiles 2004-07-21T15:28:18Z fwiw .rdf mime type is in next apache mime types. FINALLY 2004-07-21T15:28:29Z cool! 2004-07-21T15:29:01Z In a large dir it helps but, yes, its just a name 2004-07-21T15:30:57Z we have regular test cases. these are likely to be numerous and persnickety -- not something the average joe will look at until trying out some implementation of DAWG-QL 2004-07-21T15:31:33Z there are examples from the document. for consistency, we should be able to run those as test cases. maybe they are the first test cases to get the new tester on his feet. 2004-07-21T15:32:12Z the spec could have an appendix with the test results from queries that are in the spec. 2004-07-21T15:32:18Z ericP, hmm. I'm getting "Permission denied (publickey)." from cvs and ssh 2004-07-21T15:32:41Z just a sec, need to recite a grocery list... 2004-07-21T15:34:34Z nm, I was trying swh instead of sharis2 2004-07-21T15:37:13Z whats the module name for DAWG? I cant find i listed on the main page 001/sw/DataAccess 2004-07-21T15:38:22Z thx, gotit now 2004-07-21T15:39:36Z back 2004-07-21T15:41:32Z so all three of us can check out and commite to WWW/2001/sw/DataAccess 2004-07-21T15:41:42Z i guess only i can set ACLs there 2004-07-21T15:42:18Z have either of you got a master plan for the test suite and the specs? 2004-07-21T15:42:52Z * AndyS just phased out - urgent system admin elsewhere :-( 2004-07-21T15:43:15Z My "plan" is to add tests in the same layout as currently. 2004-07-21T15:43:19Z Is that a plan? 2004-07-21T15:43:43Z maybe we should think about it for a few and send mail with proposals 2004-07-21T15:43:50Z Is there anything we can't test for the QL yet? 2004-07-21T15:44:03Z i'd like to think about this from the perspective of someone wanting to include our tests in their QL product 2004-07-21T15:44:16Z AndyS, possbily we cant test for things that return graphs 2004-07-21T15:44:17Z can't test, like features? 2004-07-21T15:44:20Z One thing - the manifest will have to have a status etc like RDF core had 2004-07-21T15:44:23Z ahh, got it 2004-07-21T15:44:23Z or I've just not seen the expression for it 2004-07-21T15:44:42Z Err - graph compare? That's the only test I do. 2004-07-21T15:45:01Z I turn result sets into a result set RDF file and call the isomorphism tester. 2004-07-21T15:45:33Z I just meant the dawg-result-*.n3 format 2004-07-21T15:45:37Z All results are recorded as a single graph - if we have multigraph returns it breaks. 2004-07-21T15:46:01Z how are graph returns indicated in the manifest? 2004-07-21T15:46:26Z Ah - I see - need to know the "type" of query. 2004-07-21T15:46:42Z yes 2004-07-21T15:46:48Z I get that from parsing the query and asking the parsed form what it is. 2004-07-21T15:46:53Z ugh 2004-07-21T15:47:22Z that means that the page generator s/w needs to track the spec... 2004-07-21T15:47:34Z Coudl have a new prop to say one of SELECT/CONSTURCT/DESCRIBE/ASK (by URI of course) 2004-07-21T15:47:44Z how about mf:result-graph? 2004-07-21T15:48:09Z I rather it was independent of the rq23 syntax 2004-07-21T15:48:20Z or qt:query-form? Its part of the action. 2004-07-21T15:48:42Z yes, that would be fine - result-form though? 2004-07-21T15:50:08Z Why? Just asking. All results are (in the manifest) single graphs 2004-07-21T15:50:21Z even "select" ones. 2004-07-21T15:50:40Z AndyS, yeah, but some are graphs that describe the results, and some are graohs that are the results 2004-07-21T15:50:47Z Or do you test differently to me? 2004-07-21T15:51:07Z I'm only thinking of display at this point, but yes, I suspect I test differently 2004-07-21T15:51:42Z Ack - display reasons. 2004-07-21T15:51:57Z So how do you test? 2004-07-21T15:52:22Z it would be nice if the manifest handling was independent of the query syntax too 2004-07-21T15:52:37Z I'm lucky to have JJCs isomorphism code available. I used to directly compare result sets but dropped the code 2004-07-21T15:52:50Z test: in the table case I build the table fomr the graph and diff it with what I got 2004-07-21T15:53:02Z btw, i'm confused about where we've gotten in the requirements. are we dealing (currently) only with aggregate results? or graphs supporting each result? 2004-07-21T15:53:22Z only aggregate IIRC 2004-07-21T15:53:49Z but we don't think that (or want to bet that) will always be the case? 2004-07-21T15:54:00Z One graph for all the results of a query. So far. 2004-07-21T15:54:46Z If a request has multiple graphs, it break the web arch principle of query=>resource (multipart MIME isn't it IMHO) 2004-07-21T15:54:50Z multi graph can be indicated with multiple mf:result assertions I think 2004-07-21T15:55:28Z Ah - yes - unless order matters (lists!!!) 2004-07-21T15:55:52Z rq23 doesnt have ORDER BY :) 2004-07-21T15:56:34Z So - is it a feature of the query or a feature of the result? qt:query-form vs qt:result-form? 2004-07-21T15:56:45Z Can one query form be viewd in different ways? 2004-07-21T15:56:55Z not sure what the "query=>resource" principle is or why it gets broken by enveloping protocols in the result 2004-07-21T15:57:03Z but happy to drop it now 2004-07-21T15:57:23Z Hmm - XML output - we need to test XML output directly. 2004-07-21T15:57:30Z AndyS, it feels like a feature of the reuslts to me 2004-07-21T15:58:01Z eg. (hypothically) would could change the smentics of DESCRIBE so that it returns a table 2004-07-21T15:58:12Z I suggest both - mark the query form and the result form for MIME-type effects 2004-07-21T15:58:57Z You've been spying!! I have been thinking on those lines as well! 2004-07-21T15:59:26Z The code isn't even in CVS yet! 2004-07-21T16:01:23Z AndyS, why mark the query form? Or do oyu just mean the synta of hte query (eg. <_:rq23>) 2004-07-21T16:01:36Z *syntax 2004-07-21T16:02:11Z Mark the form to test it parsed correctly + help display different forms differently. 2004-07-21T16:02:43Z I'm not sure I like the idea of either 2004-07-21T16:03:12Z different syntaxes, yes (eg RDQL v's rq23), but not SELECT v's DESCRIBE 2004-07-21T16:03:25Z mf:action 2004-07-21T16:03:25Z [ qt:query ; 2004-07-21T16:03:25Z qt:query-form dawg:select ; 2004-07-21T16:03:25Z qt:result-form dawg:graph ; 2004-07-21T16:03:25Z qt:data ] ; 2004-07-21T16:03:33Z example 2004-07-21T16:03:46Z actually - better to label the thing itself. 2004-07-21T16:04:08Z Why not S vs D? 2004-07-21T16:04:36Z seems redundant and open to being obsoleted by syntax changes 2004-07-21T16:04:57Z [ mf:name "dawg-query-001" ; 2004-07-21T16:04:57Z rdfs:comment 2004-07-21T16:04:57Z "Optional triples: single optional triple case." ; 2004-07-21T16:04:57Z mf:action 2004-07-21T16:04:57Z [ qt:query ; 2004-07-21T16:04:58Z qt:data ] ; 2004-07-21T16:05:00Z mf:result 2004-07-21T16:05:04Z ] 2004-07-21T16:05:10Z qt:query-form dawg:select . 2004-07-21T16:05:12Z qt:result-form dawg:graph . 2004-07-21T16:05:18Z (sorry for the C&P blast) 2004-07-21T16:05:45Z yeah, seems good. I can always ignore query-form if it anoys me :) 2004-07-21T16:05:56Z Sure - but explicit marking will enable us to know what is expected when things do change. 2004-07-21T16:06:06Z i don't see that dawg:select and "SELECT" are neccessarily spelled the same way. seems as safe to syntax changes as you can get. 2004-07-21T16:06:07Z And you can ignore it (== check it!!!!) 2004-07-21T16:06:22Z Just a defined term - suggestions? 2004-07-21T16:06:45Z Its chicken-and-egg 2004-07-21T16:07:13Z Need the label (URIs are not opaque in real life) that can eyeball 2004-07-21T16:07:32Z Or use a UUID : urn:uuid:ec137f02-27bc-11b2-8000-d0a4ca2d2835 2004-07-21T16:07:34Z :-) 2004-07-21T16:07:37Z i changed my mind, i think that dawg:select only indicates what we want dawg:graph to denote 2004-07-21T16:08:05Z * SteveH agrees 2004-07-21T16:08:36Z * AndyS not sure what that means "to denote" - just checking 2004-07-21T16:09:09Z one URI says "the product of this query will be a graph. do a graph match to check the results" 2004-07-21T16:09:38Z the other seems to say either that, or "this query starts with the string \"DESCRIBE\"." 2004-07-21T16:09:40Z Steve asked for indicators so it can be styled differently from just the manifest file. 2004-07-21T16:10:12Z AndyS, yeah, but the result set descriptor does that more clearly 2004-07-21T16:10:24Z And one query form may be executed for different outputs (RDF vs XML result set) 2004-07-21T16:11:17Z AndyS, if that changes the content (nto just format) of results I would expect it to have a different query string 2004-07-21T16:11:21Z Surely, result set descriptor is about the result file? (graph, xml, table) 2004-07-21T16:11:51Z That is not clear - it may be MIME-type which is not even in the URL 2004-07-21T16:11:55Z it seems like there may be another selector that helps us figure out what may come back from conneg, but i don't think that dawg:select is it 2004-07-21T16:12:39Z Correct - dawg:select says (*without* looking at the query) that it is a SELECT 2004-07-21T16:12:57Z my expectation is that the group will not decide between graph format and binding format based on conneg 2004-07-21T16:13:13Z * AndyS surprised. 2004-07-21T16:13:16Z conneg? 2004-07-21T16:13:21Z we may have different forms of bindings or graphs, based on conneg 2004-07-21T16:13:24Z oh, nm 2004-07-21T16:13:25Z content negotiation 2004-07-21T16:13:27Z Content negotiation 2004-07-21T16:14:24Z andy, surprised by "2004-07-21T16:12:57Z my expectation is that the group will ..." ? 2004-07-21T16:14:26Z SELECT=>RDF ; SELECT=>XML ; ASK=>token ; DESCRIBE and CONSTRUCT are graphs (only - modulo conversation earlier) 2004-07-21T16:14:28Z conneg changing more than just the synatx would be wierd IMH 2004-07-21T16:15:33Z i suspect we all agree on conneg constraints 2004-07-21T16:15:34Z Maybe - its an open matter. Note also service style access points (c.f. N3QL protocol) 2004-07-21T16:16:34Z What do we do now? Explicitly record in and out seems to me the best (not a strong opinion) 2004-07-21T16:17:24Z as long as the result set is marked I dont care 2004-07-21T16:17:43Z (I'm already using .n3, .rdf, .nt to change behavio{u}r :-) 2004-07-21T16:17:52Z +1 on result set 2004-07-21T16:18:11Z yeah, we can see what falls out. 2004-07-21T16:18:31Z for now, we can specify what the result-set property means and figure out the other. 2004-07-21T16:18:40Z I was trying to distinguish form from format form=(describe or select or ask or ...) in the ql and format is the syntax of the form, negotiated by the protocol (maybe defaulted in the ql), such as by conneg 2004-07-21T16:18:41Z AndyS, are you going to commit your manifest and test files? 2004-07-21T16:20:30Z daveb, as an impelementor, how would you like the test suite organized? 2004-07-21T16:20:37Z Yes - where? The vocas are on W3C CVS but elsewhere 2004-07-21T16:20:40Z is it sufficient that it all be in some subdir? 2004-07-21T16:20:44Z s/vocas/vocabs/ 2004-07-21T16:21:49Z As we incrementally improve the RDFS, it would be good to have a W3C namespace. 2004-07-21T16:21:57Z ((Stuff around http://www.w3.org/2003/03/rdfqr-tests/recording-query-results.html)) 2004-07-21T16:22:24Z a regularly formatted manifest in rdf. regular means I can write a quick perl parser for it, so I can drive it outside rdf-land. suffixes for implict format changes is bad. avoid N3 :) 2004-07-21T16:22:56Z I will need a BRQL CVS soon (module of Jena prob) - for the distribution 2004-07-21T16:23:00Z avoid N3? why's that? 2004-07-21T16:23:05Z (and to work at home!) 2004-07-21T16:23:09Z n3's ever changing, a different topic 2004-07-21T16:23:25Z how about turtle? 2004-07-21T16:23:31Z Not really. The RDF subset is stable. 2004-07-21T16:23:43Z then you need to say the rdf subset 2004-07-21T16:23:47Z turtle is one way to do that 2004-07-21T16:23:50Z - or - is there a useful, stable subset of N3? 2004-07-21T16:23:51Z ok 2004-07-21T16:23:52Z but rdf/xml is the re syntax 2004-07-21T16:23:54Z Turtle is fine (if you fix those issues I have!) 2004-07-21T16:24:20Z Suggest: we work in N3 for readability but the master is RDF/XML. 2004-07-21T16:24:37Z what issues? remind me by email please 2004-07-21T16:24:39Z I assume everyone has a simple N3->RDF/XML 2004-07-21T16:24:46Z AndyS, I dont 2004-07-21T16:24:48Z no 2004-07-21T16:24:51Z could you add log:conflates and log:complet-web-deductive-closure to turtle? 2004-07-21T16:25:01Z Sure. All minor IIRC but I want to freeze the syntax to fix the parser. 2004-07-21T16:25:08Z that's not in the rdf subset 2004-07-21T16:25:10Z log:undo? 2004-07-21T16:25:16Z log:dwim? 2004-07-21T16:25:33Z log:complet-web-deductive-closure is calculated by consulting everyone who can put up web pages 2004-07-21T16:25:58Z i like log:undo but don't get dwim 2004-07-21T16:26:09Z I'm having problems about how to freeze Turtle, slightly OT here 2004-07-21T16:26:14Z maybe the IETF would help 2004-07-21T16:26:41Z Shall I revise the RDFS and do the manifest (helps to do both for checking reasons)? 2004-07-21T16:27:01Z Cold turtle == cold turkey? 2004-07-21T16:27:03Z im fin with the manifest et.c being in turtle FWIW 2004-07-21T16:27:12Z s/fin/fine/ 2004-07-21T16:27:38Z (I only edit one and produce the other machanically - less errors - or the same errors anyway) 2004-07-21T16:28:03Z sure, go ahead 2004-07-21T16:28:18Z where will you commit these tests? WWW/2001/sw/DataAccess/tests/ ? 2004-07-21T16:28:49Z Namespace : good time to change and put in W3C space. In WG area or date space? 2004-07-21T16:29:49Z WG area really is date space, but i see your distinction 2004-07-21T16:29:57Z This change is back-compat so no big deal 2004-07-21T16:30:01Z ericP, where should I put the generator code? 2004-07-21T16:30:51Z contemplating WWW/2001/sw/DataAccess/tests/gen/ vs WWW/2001/sw/DataAccess/gen/ ... 2004-07-21T16:30:58Z what do you think? 2004-07-21T16:31:20Z tests/gen seems ok 2004-07-21T16:31:37Z I only noticed the other day that the rdf tests are in datespace, and define rdf/xml. So could be edited after the REC. Lol 2004-07-21T16:31:38Z may need get/ for other tings 2004-07-21T16:31:51Z does andy put is stuff in /tests/ or in /tests/???/ ? 2004-07-21T16:32:34Z tests/manifest.n3 and tests/???/*.rdf seems ok 2004-07-21T16:32:43Z incase we want to divide up the tests 2004-07-21T16:32:48Z any bids on ??? ? 2004-07-21T16:32:56Z subdirs yes please 2004-07-21T16:33:06Z just call it part1 2004-07-21T16:33:10Z iso date? 2004-07-21T16:33:37Z what word did you chose for the brql^wwhatever doc? 2004-07-21T16:33:45Z rq23 2004-07-21T16:33:45Z Not yet but I'm sure it wil be clear sometime. Part1, 2 ,3 for now but expect change to meaningful groups. 2004-07-21T16:34:32Z having these parts will keep us from having to delete web stuff if we re-org. 2004-07-21T16:34:43Z So I will put RDFS and vocabs in top, and a subdir of test right? I might as well do the shuffle. 2004-07-21T16:34:54Z (but I'm on holiday tomorrow) 2004-07-21T16:35:09Z WWW/2001/sw/DataAccess/tests/{gen,part1,part2,manifest.n3} 2004-07-21T16:35:22Z I'll take CVS home with me. 2004-07-21T16:35:32Z leave some for the rest of us 2004-07-21T16:35:44Z whats "/gen" 2004-07-21T16:36:06Z (my way of CVS setup is cloning old area) 2004-07-21T16:36:11Z should the gen/foo.pl make tests/index.html? or someting in the gen subdir? 2004-07-21T16:36:21Z andy, see 2004-07-21T16:30:01Z and following 4 lines 2004-07-21T16:36:44Z Ok - not "general" 2004-07-21T16:36:52Z steveH, thinking... 2004-07-21T16:37:23Z Makefile or driver script in tests/ 2004-07-21T16:37:42Z Makefile 2004-07-21T16:37:43Z Pass params to gen/foo.pl 2004-07-21T16:37:55Z do we want mod_dir to give a listing for this directory? 2004-07-21T16:38:08Z I like code that writes to stdout for testing so coudl also use "foo.pl > index.html" 2004-07-21T16:38:29Z AndyS, yes, thats what it does 2004-07-21T16:38:30Z No reason why not AFAICS 2004-07-21T16:38:44Z but in the end, do we want folks to possibly be able to do a directy index? 2004-07-21T16:39:04Z or is that what you just answered? 2004-07-21T16:39:07Z (andy) 2004-07-21T16:39:09Z ericP, they can get a tarball of the manifest + files, so it should be neccesary 2004-07-21T16:39:18Z *shouldn't 2004-07-21T16:40:04Z i have a mild preference for being able to walk a tree like this with my web browser (not on the local disk), but not strong. 2004-07-21T16:40:10Z I suggest "yes to index" - I can see what's changed by timestamp 2004-07-21T16:40:14Z this comes from lookign at stuff on sourceforge 2004-07-21T16:40:20Z true 2004-07-21T16:40:26Z And don't forget to zip 2004-07-21T16:40:39Z that might suggest tests/data/manifest,part1, etc 2004-07-21T16:41:13Z links to foo.html are frowed upon over foo/, right? 2004-07-21T16:41:25Z Manifest per part or manifest for whole lot? 2004-07-21T16:41:27Z i can't remember if there's a mod_index file that allows one to get a listing *and* look at an index file 2004-07-21T16:42:18Z ambiguous "dir/" surely? 2004-07-21T16:42:31Z links to foo.html eliminate potential future conneg tricks, but i have yet to use http://www.w3.org/Overview.rdf 2004-07-21T16:43:41Z For now, let's concentrate on "part1" and gloss over super manifests 2004-07-21T16:44:08Z tests/data/* is kinda neat anyway - easier to build tarballs 2004-07-21T16:44:25Z right 2004-07-21T16:44:32Z (+1) 2004-07-21T16:44:55Z Why not keep queries, data (theer are typically few of these) and results in flat dir. It is big ... 2004-07-21T16:45:15Z but subdirs are just as unwieldy 2004-07-21T16:45:47Z * AndyS does not care that much 2004-07-21T16:45:51Z I meant tests/data/{manifest.n3,schema.n3,part1} 2004-07-21T16:45:56Z data/ is bad name 2004-07-21T16:46:07Z rdf test suite had manifests per dir and made the supermanifest from them 2004-07-21T16:46:28Z ??? What does extra level help with? 2004-07-21T16:46:28Z +1 to that 2004-07-21T16:46:32Z it was handy to concenrtate on a set of tests 2004-07-21T16:46:41Z AndyS, mayking tarballs of just the .rdf stuff 2004-07-21T16:46:47Z i'm hoping we can leave room for removing data and queries from the "current tests" wihtout leaving 404s on the web. 2004-07-21T16:46:56Z that's adjusting the manifest 2004-07-21T16:46:58Z OK - got it. 2004-07-21T16:46:58Z but i guess the partn does that 2004-07-21T16:47:25Z At the moment, 404's are to be expected. Not published. 2004-07-21T16:47:37Z see /WWW/2000/10/rdf-tests/rdfcore/ 2004-07-21T16:47:45Z people may download a few things they don't need cause they've been removed from teh manifest but still exist on the web, but maybe that doesn't matter. 2004-07-21T16:47:57Z it worked out ok 2004-07-21T16:48:03Z if they wanted to test old things. shrug 2004-07-21T16:48:11Z CVS! 2004-07-21T16:48:40Z I woudl keep old tests and flag them "don't do it" in the manifest if they are worth preserving. 2004-07-21T16:49:00Z mf:status = "Old stuff" 2004-07-21T16:49:13Z someone looking in the mailing list archive may find some discussion and a link a test we've since "deleted" and 2004-07-21T16:49:33Z they are marked as such 2004-07-21T16:49:38Z go read the rdf test suite rdf files 2004-07-21T16:49:41Z get a 404 when they could ahe gotten some usefull insite to how the WG decided to proceed 2004-07-21T16:50:06Z Dangerous to keep tests that become meaningless if we do this test-driven-development style 2004-07-21T16:50:16Z encourages people to try to preserve old stuff 2004-07-21T16:50:21Z as long as you direct things via the manifest 2004-07-21T16:50:26Z flagging old tests and not deleting them -- sounds good to me 2004-07-21T16:50:33Z yep, me too 2004-07-21T16:50:49Z we can add comments to the query explaing why its not used 2004-07-21T16:50:51Z you can always add a big scary header in the obsolete test commnets, for people to read 2004-07-21T16:51:33Z does this remove our need fro the extra tests/data subdir? 2004-07-21T16:51:58Z there is no need - jsut makes distribution easier 2004-07-21T16:52:05Z and browsing 2004-07-21T16:52:23Z ok, happy with it then 2004-07-21T16:52:25Z andy? 2004-07-21T16:52:27Z dave? 2004-07-21T16:52:59Z I don't mind about the data 2004-07-21T16:53:06Z Ack. 2004-07-21T16:53:18Z Happy with data/ (better name?) 2004-07-21T16:53:25Z better name would be good 2004-07-21T16:53:29Z Keep .html from the rest 2004-07-21T16:54:05Z What about gen/ => tools/ 2004-07-21T16:54:09Z don't call them inputs/outputs in the filenames since several rdf tests got burnt that way, the data files were used in different ways by different tests. 2004-07-21T16:54:20Z AndyS, yes tools is good 2004-07-21T16:54:21Z "got burnt" -. nearly 2004-07-21T16:54:40Z tx for the warning 2004-07-21T16:54:48Z Let me sketch a summary ... give me 2 mins and I will paste into IRC 2004-07-21T16:55:01Z * DaveB sketches plans for food 2004-07-21T16:55:01Z ... 2004-07-21T16:56:43Z Add new properties: 2004-07-21T16:56:43Z mf:status 2004-07-21T16:56:43Z qt:result-form 2004-07-21T16:56:43Z qt:query-form 2004-07-21T16:56:46Z Namespace to DataAccess area. 2004-07-21T16:56:50Z Dirs: 2004-07-21T16:56:54Z tests/ 2004-07-21T16:56:56Z Makefile or script 2004-07-21T16:56:58Z tests/data/ 2004-07-21T16:57:00Z tests/data/manifest.rdf 2004-07-21T16:57:06Z tests/data/part1/ 2004-07-21T16:57:12Z tests/data/gen/ 2004-07-21T16:57:16Z tests/data/part1/manifest.rdf 2004-07-21T16:57:18Z tests/data/part1/data-001.n3 2004-07-21T16:57:20Z tests/data/part1/query-001.n3 2004-07-21T16:57:22Z tests/data/part1/result-001.n3 2004-07-21T16:57:26Z Andy will do RDFS ASAP 2004-07-21T16:57:30Z s/gen/tools/ 2004-07-21T16:57:35Z s/gen/tools/ ? 2004-07-21T16:57:40Z ack 2004-07-21T16:57:52Z See 16:54 2004-07-21T16:57:52Z also tests/index.html ? 2004-07-21T16:58:00Z Good point. 2004-07-21T16:58:18Z and tests/data.(tar.gz,zip) 2004-07-21T16:58:26Z We can have mod_dir on data/ 2004-07-21T16:59:22Z rssagent, pointer 2004-07-21T16:59:32Z RRSAgent: pointer 2004-07-21T16:59:32Z See http://www.w3.org/2004/07/20-dawg-irc#T16-50-52-1 2004-07-21T16:59:54Z Is yesterday 2004-07-21T17:00:18Z 07/21 is 404 2004-07-21T17:00:22Z i can mail a log 2004-07-21T17:00:28Z No matter - I have local log 2004-07-21T17:00:36Z oh it's not logging 2004-07-21T17:00:41Z I dont, I'd like a copy please 2004-07-21T17:00:56Z but mine has this cool GMT time stamp 2004-07-21T17:00:58Z I should give ralph my xchat2rdf prog 2004-07-21T17:01:01Z will do, stege 2004-07-21T17:01:09Z Will email log and revised summary before I go 2004-07-21T17:01:13Z ericP, thx 2004-07-21T17:01:23Z do no rdf on holiday AndyS :) 2004-07-21T17:01:32Z Ok - ericP can do it :-) 2004-07-21T17:01:42Z Defn of holiday! 2004-07-21T17:01:50Z that way they get my cool GMT timestamp 2004-07-21T17:02:21Z Anyway, kids are around. And play HP2 2004-07-21T17:02:24Z (Harry Potter that is) 2004-07-21T17:02:28Z 2004-07-21T14:44:51Z Eric - I have a ssh key question later. 2004-07-21T17:02:33Z Ack 2004-07-21T17:02:35Z still relevent? 2004-07-21T17:02:54Z Can you see how many keys I have registered? 2004-07-21T17:03:16Z ... 2004-07-21T17:03:20Z OpenSSH v2 works but sometime I need to use PuTTY and keys are different (private ones) 2004-07-21T17:03:41Z so I have two key sets. Can I register a second key? 2004-07-21T17:03:46Z username? 2004-07-21T17:03:50Z aseaborne 2004-07-21T17:04:30Z 1024 3e:b7:c1:d1:aa:49:82:37:f6:fd:e7:2e:74:ff:1a:13 Andy@atlas 2004-07-21T17:04:44Z mail or paste the 2nd 2004-07-21T17:04:48Z i'll put it in 2004-07-21T17:05:34Z Ta. No rush. 2004-07-21T17:06:03Z I have one working config - its just WinCVS can only use one ssh at a time (not unreasonable IMHO) 2004-07-21T17:06:40Z and SF is best with PuTTy depsning on network config (don't ask!) 2004-07-21T17:06:40Z ok. it's easy for me to do when you have access to the other key 2004-07-21T17:07:12Z (and why can't putty use the same keys? that's probably a big pain for folks.) 2004-07-21T17:07:50Z Different private key formats IIRC - might be my misuse. I'll look at that as well. 2004-07-21T17:08:05Z I'm going now, I'l work on the generator code some more 2004-07-21T17:08:05Z So I have gen'ed different keys for different systems. 2004-07-21T17:08:12Z Have a good holiday Andy 2004-07-21T17:08:20Z -!- SteveH [swh@152.78.190.249] has left #dawg [Leaving] 2004-07-21T17:08:22Z Just one day ! 2004-07-21T17:08:29Z CU all 2004-07-21T17:08:33Z i think i've heard it's at least difficult 2004-07-21T17:01:09Z Will email log and revised summary before I go 2004-07-21T17:01:13Z ericP, thx 2004-07-21T17:01:23Z do no rdf on holiday AndyS :) 2004-07-21T17:01:32Z Ok - ericP can do it :-) 2004-07-21T17:01:42Z Defn of holiday! 2004-07-21T17:01:50Z that way they get my cool GMT timestamp