[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [xdi] XDI graph as XRIs
|
Bill, thanks, this helps advance my
understanding. However I’m curious as to how in XDI RDF we could deal
with just resources and do away with literals. Isn’t that like saying
Java could do away with primitives, or native RDF could do away with literals? How would we treat literals as resources?
Doesn’t that create a conflict between an XDI object being a reference to
a resource vs. being a literal? At the Higgins F2F meeting today we just
discussed the same topic, and they concluded it was important to keep the
distinction between whether an object is a literal or a reference to another
resource. In both the proposed X3 syntax and the XDI
RDF XML syntax, the distinction between whether an object is a literal or a
reference (or a subcontext) is crystal clear. My gut tells me to lose that
distinction would cause us grief. So I’m curious as to what you have in
mind for why we want to lose that distinction. (Unfortunately, due to YET ANOTHER CONFERENCE,
I won’t be able to attend tomorrow’s telecon to discuss this in
person, but I’ll keep monitoring the email thread. Next week my
month-long string of conferences is over and I’ll be able to attend.) =Drummond From: Barnhill,
William [USA] [mailto:barnhill_william@bah.com] Hi Nat, Re: literals... in RDF they are treated as a separate type
of entity. RDF then has two types of entities: Resources (comprised of URIs and
Blank Nodes) and Literals (typed and untyped). My proposal would make an XDI
semantic graph only have one kind: Resources. The closest analogy I can
come to is Java, which is object oriented except for primitives. Think of
literals as RDF's version of primitives, so RDF as resource-oriented except for
literals. XDI would be purely resource-oriented. Re: encoding... I figured URL encoding. As I mentioned
in an earlier email I wouldn't expect literals within XRIs to contain big
amounts of data, just like you wouldn't store a big binary file within an RDF
data store as an escaped literal. Instead you'd store it in a binary store that
provided XRI addressing for it's contents and it would cease to be a literal. Bill Bill BarnhillAssociate (XML, Emerging Technologies, Web Services, Java, Ruby)Booz | Allen | Hamiltonmailto:barnhill_william@bah.comphone:+1.315.330.7386 // +1.315.491.6765 (cell)i-name: xri://=Bill.Barnhill From: Nat
Sakimura Hi Bill,Not knowledgeable on RDF discussions, I am not quite in a position to assess the statement in 1), and would appreciate if you can explain the issue a little more so that I can get up-to-speed.For 2), what kind of encoding scheme do you propose? In case of RDF, since it essentially is an XML, we could have 0x10 0x13 in it, but for XRI, I do not think it is feasible. I guess we would have to develop an escaping scheme for line folding also.=natBarnhill, William [USA] wrote:>> Regarding the items you listed as possible disadvantages:> 1) The difference between a literal and a reference is not so clear; > literals are not first-class objects in the graph> I think this is actually a good thing, it means you can say everything > that is a first class object in the graph is an XRI. It also allows us > to circumvent some of the issues with literals that RDF has run into. > I recently had a round (or 4) with that while working with the > Semantic Web Rule Language (SWRL).>> 2) Encoding is required for literals that contain characters not > allowed in XRI syntax> Yes, and that's a pain, but is also required for literals in RDF. For > example you have to XML escape any literals with containing XML in > RDF, unless you're using RDF/XML (and even then the support for > rdf:parseType="Literal" varies.>> You brought up two good points, but because of the above reasons I > still think literals as XRIs are 'a good thing'.>> Bill> -----Original Message-----> From: Markus Sabadello [mailto:markus.sabadello@gmail.com]> Sent: Tue 5/20/2008 3:33 PM> To: Barnhill, William [USA]> Cc: Giovanni Bartolomeo; Drummond Reed; Bill Barnhill; Nat Sakimura; > tatsuki@nri.com; xdi@lists.oasis-open.org> Subject: Re: [xdi] XDI graph as XRIs>> It seems to me the advantages of your way of doing literals are:> - An XDI graph can easily be expressed as a set of XRIs> - Predicates can have more than one literal> - Subjects can be literals>> The disadvantages maybe are:> - The difference between a literal and a reference is not so clear; > literals are not first-class objects in the graph> - Encoding is required for literals that contain characters not > allowed in XRI syntax>> Also, I'm not sure what to think about having the type in the literal > instead of in the predicate.>> Markus>>> On Tue, May 20, 2008 at 10:31 AM, Barnhill, William [USA] > <barnhill_william@bah.com> wrote:>>> > Comments inline...> ________________________________>> From: Markus Sabadello> Sent: Tue 5/20/2008 12:20 PM> To: Barnhill, William [USA]> Cc: Giovanni Bartolomeo; Drummond Reed; Bill Barnhill; Nat > Sakimura; tatsuki@nri.com; xdi@lists.oasis-open.org>> Subject: Re: [xdi] XDI graph as XRIs> > > > I have seen the type in the predicate many times.> > E.g. would this> > =bill.barnhill> +email$xsd$string> "barnhill_william@bah.com"> > be the same as this?> > =bill.barnhill> +email> $xsd$string*' > <mailto:$xsd$string*%27barnhill_william@bah.com%27> > barnhill_william@bah.com'> > {wab:> Not sure, depends on what you mean by 'same', but I think the > answer to what you mean is yes.> }>> Could you even still call the second example a "literal"? > Looks like it would just be an XRI segment like everything else?> > {wab:> Yep, Exactly! That's why I said in my earlier email I prefer > the approach of not having literals at all (like Drummond says), but > still being able to capture the semantics that literals imply by > having literals just be another type of XRI. What do you think?> }>> Markus> > > On Tue, May 20, 2008 at 4:24 AM, Barnhill, William [USA] > <barnhill_william@bah.com> wrote:> >> Hope this gets through to the list.> >> Thank you Giovanni for bringing up the typing point. > For typing I've been treating 'xxx' as a subsegment that represents an > untyped literal that defaults to the string type $xsd*string (also > seen it in emails as $xsd$string, but am now liking first method > better to avoid namespace clutter).>> I'm using the idea that any literal is a name within > the namespace consisting of all literals, and a typed literal is a > name within the sub-namespace of literals that is the namespace > consisting of all literals of that type.>> '11' then translates to $xsd*string*'11'>> and the number 11 as an integer literal would be > $xsd*integer*'11'.>> Thanks,>> Bill>>> ________________________________>> From: Giovanni Bartolomeo> Sent: Tue 5/20/2008 6:08 AM> To: Markus Sabadello; Drummond Reed> Cc: Bill Barnhill; Nat Sakimura; tatsuki@nri.com; > xdi@lists.oasis-open.org> Subject: Re: [xdi] XDI graph as XRIs> > > Guys, welcome back from the Summit!> > I'm reporting some excerpts from one of Bill's mail > (it was about to allow literals as subject):> > >> [..] Post is long so won't reproduce it here, > but I strongly urge a read for anyone interested in this issue. To me > the arguments in that post, and those made on the list here, put me > firmly on the side of allowing literals as subjects.> > But it poses a further question: How do we > represent them? Do we allow 24 and "xx" as literals or only "24" and > "xx"? Furthermore if all subjects are XRIs (and I think they should > be), how is a literal an XRI?> > What is a literal? It could be viewed as a > name in the namespace context of it's type. So if we pick one of the > following we can treat all literals as XRIs:> a. There is an open class of dollar words that > start with $" and end with ", e.g. $"24", $"foo" that represent > untyped literals, and you represent a typed literal via either > $xsd.int <http://xsd.int/> *$"24" or $"24"/$isa/$xsd.int <http://xsd.int/>> b. The class of literals is a special class of > xrefs that begin with a quote after the opening paren and end with a > quot before the closing paren. This has the benefit of making all > literals relative to context, but detriment of making typing require $isa.>>> if we follow this suggestion, any literal will have a > counterpart represented by a valid XRI, thus storing as a set of XRIs > will be possible (other than allowing literals as subjects) - on the > other hand, I've starting investigating SPARQ, well, they have special > symbols which operates on literals, e.g.> > > "cat"@en //the literal "cat" has a counterpart> in the English language which points to a real world > entity (an animal)> "42"^^xsd:integer //the literal "42" is a integer> number!> "abc"^^dt:specialDatatype //"abc" is a special> datatype> > > so it seems that their use of "literals" is a bit more > evolved that simply storing a value.> Hope this could help a bit!> Giovanni> > At 10.02 19/05/2008, Markus Sabadello wrote:> >> These are two different topics:> > 1) Addressing - This is quite clear. > Everything in the XDI graph has an XRI address. Since a predicate can > not have more than one literal, it is sufficient to have the subject > and predicate in the XRI address, e.g. =markus/+email.> > 2) Storing the whole graph data (including > literals) as a set of XRIs - Bill says this is possible. And this is > what my question (and I think Nat's too) was about.> > Markus> > > On Sun, May 18, 2008 at 10:50 PM, Drummond > Reed <mailto:drummond.reed@cordance.net> wrote:> >>> Markus,> > > > > > This > subject has indeed come up several times before. I know Bill has > suggested that, when looked at from an RDF graph standpoint, every XDI > document can be represented as the set of RDF statements that appear > in the document. This would include all those whose object is a literal.> > > > > > > However when we refer to "the set of XRIs" represented by an XDI RDF > document, I have proposed that if the object of an XDI RDF statement > is a literal, the literal is NOT part of the XRI. In other words, if > you have the XDI RDF statement$B!D(B> > > > > > =markus> > +email> > > "mailto:markus.sabadello@gmail.com"> >>> > > > $B!D(Bthe XRI that identifies the literal object of > this statement (using direct concatenation syntax) is:> > > > > > =markus/+email> > > > > > That's as far as we've gone discussing it.> > > > > > Is there any reason that rule will not work?> > > > > > =Drummond> > > > > ________________________________>> > From: Markus Sabadello [mailto:markus.sabadello@gmail.com]> > Sent: Friday, May 16, 2008 6:12 PM> > > To: Bill Barnhill> > Cc: Nat Sakimura; tatsuki@nri.com; xdi@lists.oasis-open.org> > Subject: [xdi] XDI graph as XRIs> > > > > > Hey Bill,> > > > I'm sitting together with Nat and Tatsuki, talking > about various XDI topics. One issue that came up was the following: I > think it was you who suggested a few times that every XDI document can > be expressed as a simple list of XRIs, right?> > > For example, if I have this XDI graph:> > > =markus> > +friend> > =bill.barnhill> > =drummond> > > I could just express it using these XRIs:> > > =markus/+friend/=bill.barnhill> > =markus/+friend/=drummond> > > Right?> > > Now the question is, how does that work with literals? > If I have this:> > > =markus> > +email> > "mailto:markus.sabadello@gmail.com"> > > Then what's the XRI that represents this statement? I'm sure > someone has thought about that before, but I don't really remember how > it works or if it works at all?> > > Markus> >>>>>>-- Nat Sakimura (=nat)Nomura Research Institute, Ltd. XDI.ORG Vice Chair |
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]