Experimental IRC log swig-2009-05-22

Available formats: content-negotiated html turtle (see SIOC for the vocabulary)

Back to channel and daily index: content-negotiated html turtle

These logs are provided as an experiment in indexing discussions using IRCHub.py, Irc2RDF.hs, and SIOC.

05:46:29<mhausenblas>morning Web of Data
05:46:29<phenny>mhausenblas: 21 May 20:01Z <kasei> tell mhausenblas email or twitter is the best way to reliably ping me. very interested to talk about void2.
05:46:31<phenny>mhausenblas: 21 May 20:33Z <LeeF> tell mhausenblas be happy to talk, who is "we" that has restarted work?
05:49:43<mhausenblas>ok, thanks, both to kasei and LeeF. we is cygri, kwijibo, Jun, and myself (same four people that initiated voiD and released the first version)
06:44:51<LeeF>ACTION wonders if there are any simple tools for loading up / displaying SKOS data in the same way there are for RDF Schema (specgen, neologism, etc.)
06:45:18<LeeF>ACTION looking for a lightweight way to check that some SKOS output is reasonable, among other things
07:15:54<mhausenblas>LeeF, did you check out if http://poolparty.punkt.at/ re SKOS?
08:38:30<tobyink>I have an httpRange-14 question.
08:38:30<phenny>tobyink: 21 May 23:46Z <idmclean> tell tobyink check out prototype theory, frame semantics, mental spaces, construction grammars, word grammars, and cognitive grammars. Also, The chomsky hierarchy is a dialect of a language.
08:39:18<tobyink>purl.org is known to return 302 redirects instead of 303 redirects. This goes against TAG advice for non-information resources.
08:39:56<tobyink>(And, yes, I know the purl.org upgrade which is taking a laughably long time, will somewhat improve this.)
08:40:47<tobyink>But, if purl.org 302 redirects to a URI which in turn 303 redirects to an information resource, is that acceptable?
08:44:00<mhausenblas>tobyink: as you know, 303 is a hack. that having said, IMHO the situation you describe seems ok to me
08:44:37<mhausenblas>ACTION notes: not 303 in general, but the purpose we use it for in linked data ;)
08:44:45<tobyink>I wouldn't say 303 is really a hack.
08:45:03<mhausenblas>hm. how else would you call it then? :)
08:45:30<tobyink>Say you had a URI representing Michael Hausenblas which doesn't have a hash sign in it.
08:46:13<tobyink>Now, in an ideal world, if I did an HTTP request for that URI, I'd expect an HTTP/1.1 200 OK message and a knock on the door.
08:46:22<mhausenblas>we don't have means to distinguish information resources from entities not on the Web, and assuming we want HTTP URIs only, this is one way to flag it (the only so far that pretty much works with the AWWW)
08:46:38<mhausenblas>and?
08:47:26<tobyink>But given that the server has no way of delivering Michael Hausenblas to me in person (yet), it seems sensible for it to respond with a 303 message to see another resource.
08:47:48<mhausenblas>again, this is just because we're all used to it already ;)
08:47:50<tobyink>Once the technology exists for Michael Hausenblas to be sent over HTTP, then it will be able to respond with 200 OK.
08:51:59<mhausenblas>tobyink: take for example linked data with RDFa. there this issue simple doesn't exist.
08:53:06<tobyink>I don't think the issue occurs any less with RDFa than it does with RDF/XML.
08:53:37<mhausenblas>well, you don't have 303 in RDFa, right?
08:53:55<mhausenblas>there is no need to do any sort of redirect, etc.
08:55:09<tobyink>There's no *need* to do a redirect with RDF/XML.
08:55:27<tobyink>The need for 303 redirects occurs because of the use of slash namespaces.
08:55:38<mhausenblas>yes
08:57:00<tobyink>RDFa doesn't seem to encourage hashes or discourage slashes any more than RDF/XML does.
08:58:09<tobyink>e.g. my station data uses a slash URIs. <http://ontologi.es/rail/stations/gb/EUS> represents Euston station and does a 303 to <EUS.xhtml>.
08:59:37<mhausenblas>ok, but still, I don't see how that relates to your original question
09:00:00<tobyink>It doesn't.
09:00:05<mhausenblas>hehe
09:00:14<tobyink>I'm not using purl.org for that data.
09:00:42<mhausenblas>no seriously, I was under the impression you wanted to hear what people think about the situation you describe
09:01:00<mhausenblas>and that where my 2c, sorry ;)
09:04:02<tobyink>About url1->302->uri2->303->uri3 ? Yes, that's a hack. I don't think 303 in general is though.
09:06:02<mhausenblas>ACTION still thinks 303 is a hack as it overloads an HTTP status code (contrast that with eg MGET)
09:06:22<mhausenblas>now, I'm not saying that I think we should invent new things here
09:06:34<mhausenblas>I indeed use and ack that 303 works
09:07:34<mhausenblas>however, tobyink, in terms of a technical 'clean' solution, MGET is, and 303 is not, IMHO
09:08:05<mhausenblas>the costs, however, to introduce MGET are too high, hence we live and work with 303, which is fine for me
09:08:36<mhausenblas>(you note the subtle difference between clean and social/businuess costs?)
09:08:45<tobyink>If you're requesting a URI which represents a resource which a server can't (or won't) deliver, the server has two options - respond with a 4xx error or suggest that the client tries a different resource. That makes sense to me anyway.
09:09:02<mhausenblas>ok, I give you another example
09:09:35<mhausenblas>I'm working in the W3C Media Fragments WG and there we need to decide now what to return for an empty segment
09:09:54<mhausenblas>(just to get away from 303; this is a more general problem)
09:10:19<mhausenblas>now, it seemed clear to us in the first round that a 4xx class (but not 404) should be returned
09:10:43<mhausenblas>(see http://www.w3.org/2008/WebVideo/Fragments/wiki/TestCasesDiscussion)
09:11:15<mhausenblas>however, a colleague recently argued: this is a situation where the client has not erred and we should use 204
09:11:22<mhausenblas>what would you decide?
09:12:45<tobyink>204 seems sensible to me. Not sure what you mean by an empty segment though? The client has say requested a segment of video with a duration of 0?
09:13:20<mhausenblas>for example http://www.w3.org/2008/WebVideo/Fragments/wiki/TestCases#TC0002:_empty_time_segment_-_npt
09:15:08<tobyink>Then yes, I'd say 204, or a 200 which meets the request's needs (e.g. a video file with correct headers, etc, but no frames).
09:15:56<mhausenblas>yes, tobyink, this would be one line of thought
09:16:33<mhausenblas>however, now we take into account that (major) sites out there already use some parts of the Media Fragments syntax (such as Google, YouTube, etc.)
09:16:49<mhausenblas>and we need to ask: what is the default behaviour, there
09:17:22<mhausenblas>what would people expect? how would a non-confirming UA react, how should a MF-conforming UA react?
09:17:35<mhausenblas>what are the social costs that come with it, etc
09:18:18<mhausenblas>(put in other words: if we manage to choose the wrong code and break existing stuff, very unlikely people will be willing to follow the standard)
09:19:38<mhausenblas>ACTION thinks he has scared away tobyink ;)
09:21:14<mhausenblas>ha, IvanHerman, I was just thinking of you!
09:21:22<IvanHerman>:-)
09:22:05<mhausenblas>just some heads-up: Axel request change of AC rep ...
09:22:13<IvanHerman>I have seen it
09:22:49<mhausenblas>ah, didn't notice the ac-update in CC, right
09:23:15<mhausenblas>ACTION tends to forget that IvanHerman is always well informed :D
09:29:29<mhausenblas>ACTION BRB, grab some coffee
10:01:47<mhausenblas>OT - /me wants to move from netvibes to some local solutions (a nice Mac desktop app?) ... any proposals?
10:13:32<shellac>mhausenblas: I guess dashboard is the obvious one. I think there's a tool to keep them visible
10:22:46<mhausenblas>ah, thanks shellac, gonna check it out
10:25:50<shellac>see also Yahoo Widgets (formerly Konfabulator) and opera supports widgets too (*)
10:26:19<shellac>(*) Whether any of these are interoperable is beyond my ken
12:39:04<idmclean>Morning world of data.
13:41:16<tobyink>.seen melvster
13:41:16<phenny>tobyink: I last saw melvster at 2009-05-16 11:43:48 UTC on #swig
13:41:28<tobyink>.seen melvster_
13:41:28<phenny>tobyink: I last saw melvster_ at 2009-05-20 09:23:24 UTC on #swig
13:49:54<idmclean>Hi tobyink, what's up?
13:50:21<tobyink>Hello idmclean, the sky.
13:56:06<LeeF>mhausenblas, I did briefly, but the Web site didn't convince me to fill out the information for a demo account :-)
13:56:44<yvesr>is something in the lines of
13:56:49<yvesr>:this dc:description "This is a thing"^^xsd:string@en valid rdf?
13:57:13<yvesr>or should lang information not be applied to typed literals?
13:58:19<LeeF>yvesr, it's not valid - a literal can have a language tag (plain literal) or a datatype (typed literal), but not both
13:59:44<yvesr>LeeF: great, thanks :)
14:11:46<iand>ACTION waves to Cloud_
14:17:27<CaptSolo>hi iand
14:17:42<CaptSolo>Cloud_ rarely talks on IRC these days
14:20:00<idmclean>tobyink, what do we need to do with the ontology?
14:41:10<danbri_>ericP, are the old Annotea servers still live and running?
14:41:41<danbri_>at http://annotest.w3.org/access is seemed i could sign up... though the confirmation mail didn't come through (yet?)
14:46:54<tobyink>idmclean: As far as I'm concerned, http://ontologi.es/lang/core works pretty well as it is, but if there's any other features you want in it, let me know.
14:49:30<idmclean>tobyink, I've been reading about linguistics recently. I started looking down the branch of cognitive linguistics and it seems like the models and methods presented in that field would be better suited for aLoL. I think we need to add a something for describing grammar in a dialect, and describing semantics in a dialect.
14:56:13<tobyink>Right now, the core ontology has mechanisms for specifying that a particular dialect is defined by particular features - those features could be a particular syntax (grammatical) or particular semantics (or anything else really), but the ontology doesn't offer much in the way of describing the features - just saying that they exist.
15:04:20<kurtjx>a bit of a nb question about the dc namespace
15:04:25<kurtjx>basically which should i use
15:04:26<idmclean>tobyink, could you direct me to an example of an ontology which describes things like that? I want to define a couple big brush methods of describing languages using the ontology, so we can get closer to helping Anchakor template code for translation.
15:04:28<kurtjx>http://purl.org/dc/terms/
15:04:42<kurtjx>of http://purl.org/dv/elements/1.1/
15:04:56<kurtjx>or
15:06:02<tobyink>I'm more of a fan of dc/terms.
15:06:25<kurtjx>yes seems cleaner, but are they equivalent?
15:06:25<tobyink>You can't use them interchangably though. There are key differences.
15:06:31<kurtjx>right
15:06:39<kurtjx>ok
15:06:47<kurtjx>so basically pick one and stick w/ it
15:07:04<tobyink>e.g. "creator" in dcterms is a resource, whereas in dc 1.1 it's a string.
15:07:04<kurtjx>seems elements/1.1/ is more popular
15:07:37<fgiasson>not more popular, only older
15:07:41<tobyink>dc 1.1 is older, so there's a lot of legacy content in it, and a lot of people still use it for new content out of habit.
15:08:28<fgiasson>dcterms is now the core vocab used by DCMI for only a couple of months
15:08:51<kurtjx>ok, so generally /terms/ is more advanced and i should probably use this in a new data set
15:09:01<tobyink>The Dublin Core Metadata Initiate have deprecated dc 1.1 though.
15:09:08<tobyink>A:|DC Terms
15:09:39<kurtjx>thnx tobyink fgiasson
15:10:02<tobyink>A: [http://dublincore.org/documents/dces/|advice] - "implementers are encouraged to use the semantically more precise dcterms: properties"
15:10:42<fgiasson>kurtjx: sure; anyway it is not a big deal since the relationship (linkage) between DC and DCTERMS is done by DCMI to make sure you can still reason over both vocab
15:11:57<kurtjx>cool
15:12:21<kurtjx>but alas it would still break my sparql query if expecting one and getting the other
15:14:08<danbri>ACTION got the annotea service invite eventually
15:14:24<tobyink>kurtjx: Indeed, unless the SPARQL implementation does RDFS reasoning for you in the background.
15:14:25<danbri>seems http://annozilla.mozdev.org/index.html was updated last year, but not enough to work with my current firefox 3.1.x
15:15:26<fgiasson>kurtjx: this is where the fun begin; otherwise it would be too simple to justify our jobs :D
15:15:58<kurtjx>hehehe
15:16:07<kidehen>kurtjx: look at the mapper ontology we built in anticipation of the effects of the googlevocab, you can do something similar in this scenario too :-)
15:16:12<shellac>kurtjx: if you are loading the data try running a simple CONSTRUCT over it to do the rdfs mapping
15:16:24<fgiasson>kurtjx: but hopefully you can check things such as: http://docs.openlinksw.com/virtuoso/rdfsparqlrule.html
15:16:50<fgiasson>other stores have these capabilities too
15:17:08<kurtjx>cool, i'm using virtuoso anyway :-)
15:17:14<fgiasson>:)
15:17:43<fgiasson>now, what will justify your job is to keep these inference tables up-to-date :D
15:17:57<Count-Duckula>Hi, I'm using rdflib and i'm inputting some n3 then serialising the graph it generates back to n3, so I should get largely the same thing back out. It's doing some odd things with my URI's see: http://esw.pastebin.com/d1e5f6efe ... why is it namespacing some of my entries?
15:18:45<kurtjx>good question i never understood rdflib's ns handling
15:20:09<shellac>Count-Duckula: that looks more like you entered <h> a dc:comment
15:20:51<tobyink>Count-Duckula: It looks like it's doing all the ones that could be legally converted into qnames.
15:22:43<Count-Duckula>I don't suppose it's a big issue, but It's irritating and I could end up with hundreds of prefixes
15:23:37<shellac>the prefixes aren't a worry, but the data has been changed
15:26:53<Count-Duckula>shellac: It doesn't seem to have I can query the graph directly and "hello joe"'s subject is "http://www.foo.com/aaa/aaa/h" which is what it should be.
15:31:28<shellac>ok, I guess the serialiser must have a bug, because @prefix _16: <file:///home/justin/projects/myseeda/develop-eggs/obs.rms/docs/>. _16:h a dc:comment . isn't what went in
15:32:19<shellac>on another issue, dc:comment isn't a type
15:33:54<Count-Duckula>shellac: That might be a bug because now it's doing ..... @prefix _17: <http://www.foo.com/aaa/aaa/>. _17:h a dc:comment; which is correct, but prefixed for whatever reason.
15:34:42<Count-Duckula>shellac: I know, i was just coming up with any old rubbish for a prefix I had handy in the n3: dc
15:36:52<shellac>Understood. I had to check, just in case :-)
15:59:47<mhausenblas>http://www.snee.com/rdf/semweboverview.html
15:59:54<mhausenblas>B:| RDF, The Semantic Web, and Linked Data
16:00:04<mhausenblas>B: by Bob DuCharme
16:00:53<Anchakor>does Bob visit this channel? would like to tell him I like his blog
16:04:50<kasei>heya mhausenblas
16:10:40<mhausenblas>hey kasei
16:10:59<mhausenblas>Anchakor: don't think so, can't remember seeing him here
16:11:10<mhausenblas>so, kasei, thanks got your message
16:11:17<kasei>sure thing
16:11:22<kasei>still want to discuss?
16:11:35<mhausenblas>sure! :)
16:11:42<mhausenblas>so, looking at http://code.google.com/p/void-impl/issues/list
16:12:13<mhausenblas>you'll find the issue, some of them we've already reviewed and selected as candidates for voiD 2.0
16:12:34<mhausenblas>now, the two main issues will be fixing the stats module and SPARQL stuff
16:12:57<mhausenblas>hence I took an action to come back to you (SPARQL WG) and ask you if you have any feature requests
16:13:08<mhausenblas>any requirements, etc.
16:13:24<kasei>ok... reading over the two top ones now...
16:13:28<mhausenblas>I'd like to have one liaison partner (possibly you?)
16:13:29<Shepard>btw mhausenblas: http://code.google.com/p/void-impl/issues/detail?id=20&can=1 is solved, can be closed
16:13:45<mhausenblas>ah, yeah, thanks Shepard
16:14:17<mhausenblas>(we didn't went through all yet, but will bring this up next time - very valuable that you tell me, thanks!)
16:14:47<mhausenblas>so, kasei, you where the one who wrote up http://lists.w3.org/Archives/Public/public-rdf-dawg/2009JanMar/0186.html
16:14:52<kasei>as far as sparql is concerned, I think a lot of stuff from DARQ isn't particularly relevant.
16:15:18<mhausenblas>so, might makes sense if we two liaise ... but dunno
16:15:19<kasei>yup, that was me (based on a rather cursory look at the existing approaches)
16:15:22<mhausenblas>right, I see
16:15:22<kidehen>mhausenblas: I take issue with purpose of 303 in Linked Data as hack. We are just using HTTP to make a protocol for disambiguating the Name and Address/Access functions of an HTTP URI
16:16:01<mhausenblas>kidehen: again, same as I said tobyink. I'm not criticising it nor do I propose something else
16:16:21<kidehen>mhausenblas: so you really mean: its a nifty / good hack.
16:16:24<mhausenblas>I'm actually using it quite actively, as you may Imagine
16:16:35<mhausenblas>how ever you interpret hack l)
16:16:40<kidehen>mhausenblas: hacks don't always imply good, sometimes they imply parochial tweaks
16:16:52<kasei>in particular, the required bindings stuff is outside the scope of sparql, and the capabilities seem overly specific w.r.t. bound predicates
16:17:06<mhausenblas>no, seriously, hack in the sense of not initially per design or so
16:17:20<mhausenblas>ok, kasei, how shall we init this discussion
16:17:52<kidehen>mhausenblas: when something is not in the initial design of something that is part of a grander design, you get a nifty hack :-)
16:17:57<mhausenblas>I mean, would you mind bringin it up in the WG and the just use the issue tracker to input your requirements?
16:18:24<mhausenblas>kidehen: actually, most of the stuff we depend on is a hack
16:18:29<kasei>can you send me an email with the google code link from above, just so i have a permanent record, and then I can respond more thoroughly?
16:18:45<mhausenblas>but when you compare it to MGET, which is *designed* to do that ...
16:18:51<mhausenblas>sure kasei, will do
16:19:05<kasei>I can bring up that we're discussing the service description stuff at the next (tuesday) sparql telecon.
16:19:11<mhausenblas>cool!
16:19:22<kidehen>mhausenblas: no, what you depend on is the power of a host environment as the infrastructure it provides. The prowess of HTTP are on display re. Linked Data and FOAF+SSL etc..
16:19:31<mhausenblas>you know, I just like to ensure that there is a single point of entry ;)
16:19:36<kasei>point people to the void issues list, look for comments, etc., and then we can try to sort it out from there...
16:19:46<Anchakor>anybody knows how is best to visualize a rdfs/owl ontology?
16:19:48<mhausenblas>yep, kasei
16:19:49<kidehen>mhausenblas: anyway, back to other stuff :-)
16:20:04<mhausenblas>kidehen: yep (Guiness, in my case ;)
16:20:18<kidehen>mhausenblas: btw - uriburner.com now has a proper home page etc..
16:20:24<mhausenblas>(it's Friday, our after-work meeting starting soon ;)
16:20:32<mhausenblas>very cool, kidehen !
16:20:57<kidehen>mhausenblas: we are now in the next stage re. this service. The final step is putting the V6 Cluster Edition behind it etc..
16:21:05<mhausenblas>(and very nice looking indeed, kidehen )
16:21:14<kidehen>mhausenblas: thnxs
16:21:51<mhausenblas>kidehen: thank you! :)
16:21:54<mhausenblas>ACTION BRB
16:27:17<mhausenblas>ACTION sent mail, kasei
16:28:23<kasei>thanks
16:32:08<shellac>mhausenblas: MGET never felt great. the link header stuff is much more appealing to me.
16:36:50<PovAddict>MGET?
16:37:06<PovAddict>metadata-get?
16:37:44<PovAddict>ah, multi-get
16:38:33<shellac>PovAddict, you were right first time
16:38:43<PovAddict>well I found http://1997.webhistory.org/www.lists/www-talk.1994q4/0479.html
16:38:53<shellac> http://sw.nokia.com/uriqa/URIQA.html
16:39:59<PovAddict>chump dead?
16:40:17<Shepard>no, it's the space in front of it
16:40:23<shellac>I sneakily added a space
16:40:25<PovAddict>ah i dont see it here
17:05:51<kasei>that's what I've done, but I didn't find it terribly useful.
17:06:08<kasei>trying to use multiple channels in one window, but can't figure out how to switch between them.
17:06:34<kasei>I can only find docs on how to switch between windows, not channels.
17:09:20<Anchakor>kasei: alt+arrows/numbers
17:10:29<kasei>that worked for windows, but doesn't seem to be working for channels :(
17:11:47<Anchakor>well I got window = channel and thats fine by me
17:13:34<mattl>kasei: alt+a is the best one
17:14:05<kasei>alt, or meta?
17:14:17<kasei>Anchakor: thanks anyway
17:14:40<kasei>meta-a doesn't seem to work either
17:16:00<mattl>kasei: which OS are you using? which terminal?
17:17:29<kasei>os x + terminal + screen + ssh + irssi on debian
17:18:04<mattl>okay, and do you have the alt = meta option set in os x terminal?
17:18:06<Anchakor>kasei: I use meta for controlling the WM
17:18:29<kasei>no, but esc is working as meta for everything else, and I'm fine with that.
17:18:54<mattl>you might need to check your screenrc
17:19:28<kasei>for what?
17:19:51<mattl>to allow a different key binding.
17:19:58<kasei>meta is working in irssi for switching between windows, so afaict it's not a key binding issue.
17:20:26<mattl>but meta-a does nothing?
17:20:41<kasei>not that I can see
17:21:20<kasei>since I remain in #swig
17:23:11<mattl>okay, i'll /msg you, and try it when you see some activity at the bottom of the screen
17:24:45<kasei>i can see your message, but can't seem to switch to a privchat.
17:25:16<bblfish>anyone up on Description Logic notation?
17:25:26<bblfish>or is there a good irc group for that?
17:25:35<mattl>kasei: what does /win <num> do?
17:26:47<kasei>nothing, but i would assume that's expected since i've turned windows off
17:27:04<bblfish>
17:27:04<bblfish>I am trying to understand the meaning of ∀¬P.¬O
17:28:01<bblfish>that should be the set of all Objects that have all relations that are not P, to the set not O, right?
17:30:08<kasei>mattl: thanks for the help, but I've got to run to a meeting. will try to sort it out later, I guess.
17:45:00<mhausenblas>good nite Web of Data
18:48:53<bijan>bblfish: that's not typical dl notation
18:49:00<bijan>You usually can't negate the P
18:49:16<bijan>You can encode that in OWL 2 using disjoint properties
18:49:38<bblfish>yes, I found it in http://sunsite.informatik.rwth-aachen.de/Publications/CEUR-WS/Vol-447/paper3.pdf section 2.2
18:49:52<bijan>whcih page?
18:50:14<bijan>That being said, your reading isn't quite right.
18:50:31<bblfish>page 4
18:50:57<bijan>Sectoin 2.2?
18:51:05<bijan>I see it
18:51:14<bijan>OMG!
18:51:27<bijan>Ok, that's totally wacked
18:51:29<bijan>
18:51:29<bijan>P.O
18:51:29<bijan>as
18:51:29<bijan>∀O.P
18:51:31<bijan>er
18:51:44<bblfish>yes, I could not understand that notation
18:51:50<bijan>Flipping the P and the O like that basically makes a total hash of the dl notation
18:51:56<bijan>It's completely bogus
18:51:57<PovAddict>I see a square
18:52:08<bblfish>yes, that's kind of the problem I had with ∀O.P
18:52:27<bijan>Wow.
18:52:34<bblfish>but one could say, ok well give them the notation. Still I thought I should understand what it is they are asaying
18:52:54<bijan>Well, it's not clear that the notation is coherent.
18:53:05<bijan>Without a grammar, who knows?
18:53:10<bijan>But if I were reading ∀¬P.¬O
18:53:17<bijan>In "normal" dlish ways
18:53:24<bijan>(thought as a syntax expression)
18:53:45<bijan>I would say, "all the objects for which all their non-P edges are in not O"
18:54:27<bblfish>so if P is say writeAccess
18:54:49<bijan>So, for example, if R and P are disjoint, <a,b>R and b:¬O (assuming no other edges) pus a in that class
18:54:59<bblfish>that would be all the people whose non writeAccess relations are not that set of files
18:55:08<bijan>Yeah
18:55:14<bijan>Kinda nasty :)
18:55:53<bijan>But my confience is shaky because 1) I'm tired and 2) people screwing with standard notation confuse me :)
18:56:21<bijan>Plus phrasing things negatively is often hard to understand
18:56:23<bblfish>so you have to limit yourself to some domain of relations I suppose. Because you could always find some relation that relates a person and that file
18:56:35<bijan>This is always true
18:56:46<bijan>Negation is like that :)
18:57:07<bijan>This is why disjoint properties are easier to understand that negated properties
18:57:16<bijan>(Actually true for disjoint classes, I think)
18:57:47<bijan>That doesn't look like a superawesome paper to me
18:58:06<bblfish>ah, pitty, because that would be a really useful paper to have for my work
18:58:38<bblfish>they are presentating it at ESWC
18:58:45<bblfish>on the SPOT track
18:59:16<bijan>There's lots of work on RBAC and logic
18:59:20<bijan>Including some of my own :()
18:59:31<bijan>Is there something specific on it?
18:59:39<bblfish>yes, I think I read your paper on XAML
18:59:57<bblfish>something specific on what?
19:00:00<bijan>Er
19:00:02<bijan>in that paper
19:00:06<bblfish>that I need?
19:00:06<bijan>That you're looking for?
19:00:08<bijan>Yeah
19:00:39<bblfish>ah well I was hoping we could use that as a basis for http://esw.w3.org/topic/WebAccessControl
19:01:05<bblfish>which would be nice for foaf+ssl to have
19:02:31<bijan>Do you need a logic at all?
19:02:41<bijan>Why not just use XACML or some other ACL
19:03:07<bijan>Or do you need to pull in complciated role heirarchies
19:03:19<bblfish>just exploring the space... We need some way to say that a resource can be written by a group of users (my friends say)
19:04:11<bblfish>so when I publish a resource I can say, <me> see:also <https://.../me> . and <https:// .../me> can be read by my friends .
19:04:29<bblfish>so if a friend of mine can't read it, he can tell me it's a bug
19:04:45<bblfish>or if someone is not my friend, he knows he has to be nice to me to read the stuff
19:04:53<bijan>As long as it's pretty much lists I don't think you need much of anything
19:05:10<bijan>It's only if you want to express complex conditions that things become an issue
19:05:22<bijan>And, frankly, for robustness in this sort of things, you don't want that
19:05:38<bijan>I.e., most people can just manage white and black lists
19:05:43<bijan>Descriptions...er...not so much
19:06:14<bblfish>well I was hoping something that could be simple, and could also be proven to evolve. I was hoping this paper did that
19:06:21<bblfish>I think it has relations such as
19:06:49<bblfish>GroupOfPeople hasReadAccess ClassOfResources .
19:06:56<bblfish>as the base
19:07:08<bblfish>then I got stuck on the above notation
19:07:30<bblfish> In Rules 6 and 12, we abbreviate ∀¬P.¬O as ∀O.P , which allows us to assign a permission P to all ob jects in O.
19:07:47<bijan>Well, my default is to be wary
19:08:10<bblfish>will you be at ESWC?
19:08:12<bijan>The main benefit of logics in this area, imho, is whenyou have complicated stuff to model and maintain
19:08:23<bijan>The inference services can 8reallY* help there
19:08:32<bijan>But here, the people using it aren't going to be sophisticated
19:08:34<bijan>No, alas
19:08:46<bijan>For your bit, I would jsut build some custom syntax
19:08:54<bijan>Maybe off robots.txt
19:09:01<bijan>Since it already has some wildcarding
19:09:56<bijan>The closest thing to this that I know of is the PAW project
19:10:02<bijan>with which I was briefly associated
19:10:06<bijan>Used n3 rules
19:10:11<bijan>i don't know where they ended up
19:10:20<bijan>But it really felt like wrongkill for me
19:10:26<bijan>Simultaneously over and under kill :)
19:10:51<bijan>http://www.policyawareweb.org/
19:11:04<bblfish>I found a link here http://www.w3.org/2000/10/swap/doc/plans.html
19:11:17<bijan>C:|Policy Aware Web
19:11:28<bijan>Of course they were aiming for more complicated stuff (e.g., proof exchange)
19:13:11<bblfish>phew, ... well I wonder how much of this paper remains good after this notational odity.
19:13:48<bijan>Indeed
19:14:09<bijan>Almost makes me which I had a student who would have to read it and figure that out :)
19:15:02<bblfish>well I need to read it, so I can be that student, if I can get through it...
19:15:22<bijan>My advice is to extract all the formulas, translate them consistently into FOL,a nd work from that
19:15:35<bblfish>ah good idea.
19:15:44<bijan>Actually, my first advice is to see if the logic actually matters
19:15:48<bijan>Which it may not :)
19:16:03<bijan>If it doesn't matter it doesn't matter if it's right
19:16:36<bblfish>yes. Well, at least I am not the only one confused about that notation - so that saves me searching through dl books...
19:16:57<bijan>Sigh. The first sentence of the abstract already grates.
19:17:16<bijan>They don't model communites *as* lightweight ontologies I hope
19:17:19<bijan>That would be odd
19:17:37<bijan>allows us to represent expressive
19:17:38<bijan>access control rules beyond the current state of the art
19:17:41<bijan>*right*
19:17:49<bijan>I so believe that.
19:17:54<bijan>I mean, it's in the abstract!!!!
19:18:22<bijan>(Seriously, I wonder how they model expressiveness.)
19:18:30<bijan>(I doubt they have a good metric at all :()
19:18:54<bijan>"""Internet business patterns
19:18:55<bijan>such as B2B, B2C, C2C are no longer high-tech terminologies but, rather, they
19:18:55<bijan>represent everyday activities involving virtually everybody from producers to end
19:18:55<bijan>customers."""
19:19:17<bijan>I wonder if "terminoloiges" in that sentence is a spelling correction of "technologies"
19:19:36<bblfish>ouch
19:20:05<bijan>Well, I'm about ready to write it off
19:20:09<bijan>Doesn't seem promising
19:20:28<bijan>That ER diagrams are magic isn't a compelling pitch to me
19:20:48<bijan>""" The first key feature of RelBAC is that its
19:20:48<bijan>access control models can be designed using entity-relationship (ER) diagrams.
19:20:49<bijan>As such, they can be seamlessly integrated into the whole system and vary ac-
19:20:49<bijan>cording to the scale of the business."""
19:20:50<bblfish>ah. I was interested in the DL piece...
19:21:18<bblfish>ouch, I better not have you look at my paper on the same track http://sunsite.informatik.rwth-aachen.de/Publications/CEUR-WS/Vol-447/paper5.pdf
19:21:58<bijan>Yeah, but, you know, if you are massively screwing up the setting of the situation, it means you're pretty likely to screw up the technical bits
19:22:06<bijan>It's a pretty common pattern in these sorts of paper
19:22:40<bijan>People build up some weird hopeful vision and then expect more magic from the ontology or the rdf to fix the rest :)
19:23:02<bijan>Consider these two sentences:
19:23:05<bijan>"The second feature, which motivates the
19:23:05<bijan>name RelBAC , is that permissions can be modeled as relations, and differently
19:23:05<bijan>from the state of the art, e.g., RBAC [2], they can be manipulated as inde-
19:23:05<bijan>pendent ob jects, thus achieving the requirements of modularity and flexibility
19:23:05<bblfish>I suppose I was hoping that they must be DL people with no industrial experience, but if they had the logic right, then that would save the rest.
19:23:05<bijan>described above. "
19:23:17<bijan>This suggests that the roles are reified
19:23:23<bijan>(Which actually is both easy to do and standard)
19:23:33<bijan>(But is in tension with them beign relations)
19:23:54<bijan>But then "This in turn allows us to
19:23:54<bijan>model permissions as Description Logic (DL) roles "
19:23:55<bijan>Sight
19:23:57<bijan>sigh
19:24:34<bijan>And they don't cite any of the papers that have attempted translations before
19:24:38<bijan>Like mine!
19:24:45<bijan>Not a good sign
19:25:02<bijan>I'm surprised this was accepted. is this research track?
19:25:36<bblfish>your paper, the one on xacml-dl?
19:25:43<bijan>Or the one on WS-Policy
19:25:56<bijan>There are others out there
19:26:40<bijan>Ooo more broken syntax: U
19:26:42<bijan>er
19:26:45<bijan>U subclassof P:o
19:26:51<bijan>That's just not remotely any DL notation
19:26:59<bblfish>oh!
19:27:04<bblfish>which page?
19:27:05<bijan>(It's also supicous that they dont' say *which* DL)
19:27:12<bijan>The one with 2.3 on it
19:27:51<bijan>Ok, now I've moved from sad and bemused to really angry
19:27:59<bblfish>Hey, if you feel like tearing into my paper, feel free: http://sunsite.informatik.rwth-aachen.de/Publications/CEUR-WS/Vol-447/paper5.pdf
19:28:03<bijan>This is total crap
19:28:04<bblfish>criticism is always helpful
19:28:13<bijan>I mean, 1-5 are just DL formulae
19:28:31<bijan>Grr.
19:28:40<bijan>They aren't "policies" in any meaningful sense
19:29:45<bijan>Oh, man.
19:29:46<bijan>Sigh
19:29:57<bijan>If you'd like my considered advice
19:30:03<bijan>I wouldnt' even bother reading that paper
19:30:14<bijan>much less trying to translate all the formulae into fol
19:31:03<bijan>I would be prepared to defend the proposition that it literally has no content
19:31:29<bblfish>looks really nice with those DL formulae though.
19:31:42<bijan>Unless you know what they *should* look like :)
19:32:19<bblfish>I need to learn more about DL logic. I had hoped this was going to be an easy way into it.
19:32:29<bijan>Ok, well, one problem I would have with your paper is that I don't believe in web archtiecture :)
19:32:35<bijan>Try our ws-policy paper
19:32:39<bijan>It's pretty accessible
19:33:36<bblfish>well foaf+ssl seems to work, we have a number of implementations for it
19:33:46<bijan>To see your paper through my eyes, apply puppization: http://clarkparsia.com/weblog/2008/01/15/tag-youre-not-it/
19:33:49<bijan>Lots of things work
19:34:00<bijan>Working and having the explanation of why it works are different things
19:35:35<bblfish>well, that's what I was trying to work out in that paper - they why it works
19:35:35<bijan>A structural point: there's a lot of background in this paper
19:35:51<bblfish>yes, we had a few people with no background in semweb,
19:36:06<bijan>But you are publishing at ESWC
19:36:11<bijan>Audience is the dominante consideration
19:36:23<bijan>If the ratio of background to new content is too high, that's a paper smell
19:36:28<bijan>(in analogy with code smells)
19:36:35<bblfish>yes, semweb people don't have necessarily security background
19:36:50<bijan>Sure, but now you've unioned them :)
19:36:57<bijan>That obscures
19:37:00<bijan>It's a smell
19:37:03<bijan>It could be ok
19:37:07<bblfish>I agree, not optimal
19:37:07<bijan>It could be a big problem
19:38:13<bblfish>perhpas the piece of interest really starts at "SSL Authentication" part
19:38:29<bijan>I've skipped to 3
19:38:32<bblfish>that attemps to proove how SSL Authentication works flogically
19:38:45<bijan>Oh, then that's probably not background
19:38:55<bijan>If you are trying to do formal verification
19:38:58<bijan>And it's novel
19:39:08<bijan>Then it should be busted out
19:39:24<bblfish>busted out?
19:39:43<bijan>I.e., made a top level section
19:39:47<bijan>It's not background really
19:39:54<bblfish>ok
19:39:55<bijan>Let me see if I've got the high level picture right
19:40:37<bijan>You are basically comparing one protocol (SSL with a PKI trust model) with a new protocol FOAF-SSL
19:40:51<bijan>You believe there are deficionciese in the former
19:40:54<bijan>and the latter rectifies them
19:41:02<bblfish>yes. The first piece was to help our security experts on board, and to test the reasoning
19:41:17<bijan>This paper almost entirely obscures this structure :)
19:41:38<bijan>I would throw most of the background out
19:41:54<bijan>I really want a methodology section
19:41:57<bblfish>:-) Now I know who to ask next time I write a paper
19:42:12<bijan>And a good sumamry of the problems with SSL and the purported advantages of FOAL+SSL
19:42:40<bblfish>yes, that would be a good way to structure things
19:43:03<bijan>And methodologically, you should definitely look at general formal methods
19:43:11<bblfish>Well as I have to present it at the conf, I can present it that way
19:43:30<bijan>If you want to claim some property of your protocol (esp. as something lacking in another) there should be models for htis
19:43:33<bijan>For other protocols
19:43:45<bijan>E.g., deadlock,livelock
19:43:53<bijan>Congestion properties
19:43:56<bijan>Scaling properties
19:44:11<bijan>Some of these things are likely proved formally
19:44:16<bijan>some may be verified experiementally
19:44:23<bblfish>it is simple: https client certs lack distribution
19:44:41<bijan>But distribution is a structural features not a desired property
19:45:12<bblfish>in short we argue it comes down to DNs not being URIs
19:45:16<bblfish>or even URLs
19:45:26<bblfish>so you don't get a web effect.
19:45:48<bijan>Yeah...well....
19:45:53<bijan>Ok, I hear puppies again :)
19:45:55<bijan>Here's why
19:46:06<bijan>"Web effect" isn't a well defined term, by and large
19:46:13<bijan>And it's very very very difficult to operationalize
19:46:23<bijan>SO it's hard to test whether you are seeing one or not
19:46:26<bblfish>well every web page can link to any other web page
19:46:28<bijan>or, for that matter, *not* seeing one
19:46:43<bblfish>because there is a global naming system
19:47:07<bijan>But again, why is "being able to link to any other web page" a desirable property?
19:47:23<bblfish>it is for distributed social networks
19:47:35<bijan>No, it really isn't
19:47:47<bijan>for example, I'd rather have a large, active, money producing districuted social network
19:47:54<bijan>Even if it can't link to any other web page
19:48:04<bblfish>sorry I had a short cut
19:48:13<bijan>In fact, if all my distributed social network can do is "link to any other web page" it's worthless
19:48:30<bijan>(This is my point: These terms are commonly used but very very wifty)
19:48:34<bblfish>no your distributed social network allows you to link to other people wherever their foaf file is
19:48:51<bijan>Er...it does?
19:48:54<bblfish>yes.
19:49:00<bijan>What's a DSN in this case?
19:49:05<bijan>I mean, it is some software?
19:49:10<bblfish>linked data
19:49:26<bijan>Hrm.
19:49:30<bijan>You keep throwing jargon at me
19:49:35<bijan>Without it being obviously relevant
19:49:47<bijan>Methodolgoically, I suggest that you don't use any jargon
19:49:48<bblfish>perhaps I lost the thread
19:50:05<bijan>So, roll back to your protocol
19:50:12<bijan>What are some useful features of protocols
19:50:21<bijan>One obvious one is that they are bandwidth light
19:50:35<bijan>All things being equal, we prefer less use of bandwidth than more
19:50:44<bijan>Similarly, we value reliability
19:50:50<bblfish>yes, so foaf+ssl wins there over openid
19:50:54<bijan>Stop
19:50:59<bijan>Don't jump to applying it to yoru case
19:51:05<bblfish>ah ok
19:51:08<bijan>Let's just get some common ground
19:51:41<bijan>Do you agree that these are the sorts of thing we 1) can show are properties of some protocols and not of others and 2) that are generally desirable?
19:52:00<bijan>And 3) can be cached out in fairly straightforward operational terms?
19:52:01<bblfish>yes, one could attack the problem that way
19:52:15<bijan>I'm just trying to make sure we're talking about the same things :)
19:52:20<bijan>In roughly the same way
19:53:12<bblfish>yes. So one could build a paper on this around those properties
19:53:48<bijan>So, for me, I would like to know (after reading your paper), 1) what (desirable) property is missing from e.g., SSL (and some details of the desirability), 2) a proof (or other evidence) that SSL lacks it, 3) a proof (or other evience) that your protocol has it
19:53:49<bblfish>So one would want to look at authentication protocols, and the advantageous features they have
19:54:23<bijan>Evidence could be a formal proof, or results of an experient, or (more weakly) a plausibility argument
19:54:32<bblfish>Ok, I think I could offer that.
19:54:35<bijan>And then I would want to know what I lose, if anything
19:54:45<bijan>Engineering usually involves tradeoffs
19:54:49<bblfish>This paper adds an explanation of why it works
19:54:57<bijan>Expalantion? not proof?
19:55:01<bijan>Not experiemental evdience
19:55:06<bblfish>well I hope it is proof
19:55:07<bblfish>sorry
19:55:21<bijan>Well, that's something to figure out :)
19:55:43<bblfish>that's what the section 3.2 is about
19:56:30<bijan>Right so 2.7 purports to offer my 2 and 3.2 purports to offer 3
19:56:54<bblfish>yes
19:56:58<bijan>Ok, 3.2 has problems qua proof
19:57:04<bijan>because it's sort of an example
19:57:06<bijan>not a proof
19:57:14<bblfish>true, I am not a logician
19:57:15<bijan>That could be a good starting point to think about a proof
19:57:19<bblfish>yes
19:57:47<bijan>(Mathematician...really a specialized sort. It's not really about logic per se)
19:58:06<bblfish>well I think we are into doxastic logic
19:58:11<bijan>Well, in general, a good thing to do is to figure out what you're trying to prove
19:58:12<bijan>perhaps
19:58:15<bblfish>(the logic of beliefs)
19:58:28<bijan>er...you can rest assured tha I know what doxastic logic is :)
19:58:45<bijan>But the big barrier is that I don't knwo what your theoremw ould be
19:59:00<bijan>E.g., Thus FOAF+SSL has property P
19:59:11<bblfish>this
19:59:37<bblfish> the reasoning of the web server S trying to authenticate a :client.
20:00:01<bijan>?
20:00:02<bblfish>S wants to know if the request is allowed access to a resource R
20:01:29<bblfish>where a resource R is allowed access to by some agent identified by a URL
20:03:02<bijan>(is that the statement?)
20:05:49<bijan>(cause that doesn't look like a theorem at all much less one of the form Protocol has property P :))
20:10:11<bijan>Heh
20:26:59<bblfish>oops, sorry the internet café lost the connection here
20:27:38<bblfish>bijan: well we are trying to prove that in the paper. If we can proove it, it should be very useful.
20:29:15<bblfish>I am sure there are all kinds of meta properties that would also be very useful to proove
20:47:43<Shepard>does anyone know if the sessions of the 2009 semantic technology conference will be recorded in any way?
22:22:40<Anchakor>is this ok? ':DataType a owl:Class , :DataType ; rdfs:comment "set of all datatypes" . '

Back to channel and daily index: content-negotiated html turtle