[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [xri] POLL: Syntax 2.0 or 2.1
In a message earlier today regarding the example XRI of "@ootao+west*steve" Drummond said he '...was implying that @ootao is in fact authoritative for it's own definition of "+west".' If that's true, then I see no reason to support the "+" namespace at all; @ootao could just as well use "@ootao*west*steve". Drummond provided an XDI RDF address (#1) of an attribute of one of his personal personas, expressed using the direct concatenation in the XRI Syntax 2.1proposal: #1: =drummond+work/+person+name+first I suggest the following (#3) as a way to express the same (or at least similar) relationships using 2.0 syntax: #3: =drummond*work/+person*name*first In neither #1 nor #3 is it clear to me where to send the resolution query for "+person", although if someone understands what to do with +person in #1, I suppose they would also understand what to do with +person in #3. Marty.Schleiff@boeing.com; CISSP Associate Technical Fellow - Cyber Identity Specialist Computing Security Infrastructure (206) 679-5933 -----Original Message----- From: Drummond Reed [mailto:drummond.reed@cordance.net] Sent: Tuesday, May 01, 2007 9:52 PM To: 'Steven Churchill'; 'Tan, William'; 'Gabe Wachob' Cc: xri@lists.oasis-open.org Subject: RE: [xri] POLL: Syntax 2.0 or 2.1 Steve (and everyone on this thread), I think we are talking past each other RE what I'm saying about the role of XRIs and XDI RDF. XDI RDF doesn't do anything to change the RDF graph model, that's why it's called XDI RDF. Everything is triples. It does, however, use a different serialization model than RDF, and the primary reason it does that is *addressing*. As Bill Barnhill so well summarized it at our f2f meeting, the XDI RDF model has three breakthrough features over conventional RDF: 1) No blank nodes. 2) Automatic reification. 3) All statements are identifiers and all identifiers are statements. All three of these features are a direct result of the core design principle of XDI (in both the ATI and RDF models): 100% addressability of all nodes in the graph. In fact, in the XDI RDF model, not just all nodes addressable, but all XDI statements (arcs) are addressable -- even the refs (in the XDI ATI model, we needed a workaround to make refs addressable). In short, THE killer feature of XDI RDF is addressing, which depends completely on XRIs. In particular, it depends on the ability of XRI to compose identifiers out of other identifiers -- both global and local identifiers. When it comes to addressing, the value of direct concatenation to the XDI RDF model is the simplicity of the XRIs used to address nodes in the XDI RDF graph. In response to Gabe's earlier note about "having these requirements in front of us", I have repeatedly tried to make sure everyone on the XRI TC was aware of how significant this factor is by pointing them to the very detailed XDI RDF proposal, currently in its fourth version: http://www.oasis-open.org/committees/download.php/23161/xdi-rdf-model-v4 .pdf This is the XDI addressing model that I have been referring to throughout the development of XRI Syntax 2.1. If you have not read this document, I strongly urge you to take the time to read through it and look over the example XDI documents (in four serialization formats, no less) and the XDI addresses they contain. I think they provide very rich examples of the dramatic difference that direct concatenation makes in the simplicity of XDI RDF addresses. By simplicity, I mean two things: 1) Human-readability and understandability, which has a huge amount to do with adoption by developers, directory administrators, and other IT personnel who (we hope) will be dealing with XDI addresses on a day to day basis (just like they deal with IP addresses, domain names, and URIs on a daily basis), and for whom human-readability and understandability is a serious factor in adoption (witness HTML). 2) Comparability, i.e., the ability to take two XRIs and compare them without complex normalization rules. Since this topic is so important (why else would I be typing this late into the night...), let me provide a detailed example from the document I cited. The following is an XDI RDF address of an attribute of one of my personal persons, expressed using the direct concatenation in the XRI Syntax 2.1 proposal: #1: =drummond+work/+person+name+first Now, here's the closest address that one could TRY to compose that expressed the same relationships using 2.0 syntax: #2: =drummond*(+work)/(+person*(+name*(+first))) I emphasize "TRY to compose" because the grammar expressed in the second address is NOT EQUIVALENT to the first one. That's what I was trying to explain in the f2f meeting: global-xrefs, the ABNF rule that enables direct concatenation (see http://wiki.oasis-open.org/xri/XriCd02/XriAbnf2dot1) are not a "compact" way of expressing local-xrefs, as I (and all of us) first thought. As Les puts it, they are a different grammatical construct than a local-xref, because they are a reference to a resource in a global context vs. a local context. (To use an English language analogy, this is like the difference between saying "Drummond Reed" and "Drummond's Reed". Grammatically they are very different.) That point alone is very important, but back to my original point: look at these two addresses. See how intuitive #1 is. Here at the Higgins f2f, folks who had never worked with XRI have been able to just go to the whiteboard and write one without thinking. Now, imagine them trying to do that for #2. Yes, I know you can make a case (as Wil has) that "most people will not need to understand XRI syntax". And you can also make a case (as Steve has) that "global xrefs add complexity to XRI parsing and resolution". But I hope you can understand why I am pushing back so hard on those arguments. It's not that I don't understand them -- I do. It's that I believe they don't outweigh the tremendous intuitive and expressive power of direct concatenation. Besides being brain-dead easy to read, it gives any person who knows how to type the ability to tag an XRI just by adding "+tag". To say we shouldn't do this because we didn't understand how this would possible in earlier versions of XRI syntax (which is the absolute truth -- do you think that if we had seen this in XRI 1.0 or 2.0 we would not have done it??) would be the TC equivalent of saying, "Yes, cars may be faster than horses but they are more complex and require gasoline and spew exhaust, so let's just stick with horses." That's as much as I can put in an email tonight (though I'm doubtful email will be very productive towards resolving this issue). I'll schedule it for Thursday's telecon. =Drummond -----Original Message----- From: Steven Churchill [mailto:steven.churchill@xdi.org] Sent: Tuesday, May 01, 2007 6:25 PM To: 'Drummond Reed'; 'Tan, William'; 'Gabe Wachob' Cc: xri@lists.oasis-open.org Subject: RE: [xri] POLL: Syntax 2.0 or 2.1 Drummond, It strikes me as funny that you continue to use XDI RDF as the motivation for adding complexity to the abstract models of XRI syntax and authority resolution. It is ironic, because I think you have failed to take away one of the most important lessons that the RDF folks have to offer: that of the importance of a foundation based upon a solid and simple abstract model. I urge you to take a look at the normative document "RDF Concepts and Abstract Syntax" <http://www.w3.org/TR/rdf-concepts/>. If you do, you may note the following: 1. The RDF abstract graph model IS formally defined. (Kudos to them!) See section 3.1 "Graph Data Model". The XRI/XRI TCs have consistently "avoided" formally defining their abstract syntax and graph models. 2. The incredible simplicity of the RDF graph model. This is by design, not by accident. 3. Note the language of section 2.2.4: "RDF has a recommended XML serialization form [RDF-SYNTAX], which can be used to encode the data model for exchange of information among applications." To the RDF folks, the serialization format is not the "heart" of the specification -- it is a recommendation. If only RDF XDI could take away that lesson and focus on its graph model instead of its serialization format! So you keep using (XDI) RDF as a motivation for adding complexity to the XRI models. I personally consider RDF as a perfect example of why we should NOT do so. ~ Steve -----Original Message----- From: Drummond Reed [mailto:drummond.reed@cordance.net] Sent: Tuesday, May 01, 2007 3:44 PM To: 'Tan, William'; 'Gabe Wachob' Cc: xri@lists.oasis-open.org Subject: RE: [xri] POLL: Syntax 2.0 or 2.1 Again, I want to make sure it's clear why this is such a big issue: the XDI RDF model uses the "grammer" of global-xrefs. This "grammer" cannot be duplicated (at least in any easy way I can figure out) by "going back" to 2.0 syntax. So this issue is deep, deep, deep. =Drummond -----Original Message----- From: Tan, William [mailto:William.Tan@neustar.biz] Sent: Tuesday, May 01, 2007 3:33 PM To: Gabe Wachob Cc: xri@lists.oasis-open.org Subject: Re: [xri] POLL: Syntax 2.0 or 2.1 Yes. Xrefs can only appear within parentheses. Gabe Wachob wrote: > Can you spell out the implications of this choice so we know exactly > what it > is we are voting for/against? > > That is, we are saying that XREFs basically stay as is and that the > @foo+bar+baz syntax goes away? > > -Gabe > > >> -----Original Message----- >> From: Tan, William [mailto:William.Tan@neustar.biz] >> Sent: Tuesday, May 01, 2007 3:25 PM >> To: xri@lists.oasis-open.org >> Subject: [xri] POLL: Syntax 2.0 or 2.1 >> >> We did an informal poll at the f2f two weeks ago on whether we should >> stick with XRI Syntax 2.0. I suggest we vote again on the list. >> >> +1 - to support concatenated syntax >> 0 - don't care >> -1 - no concatenated syntax >> >> I'm reverting my vote to -1 because the solution is too confusing IMHO. >> >> =wil >> > > >
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]