- From: Tim Berners-Lee <timbl@w3.org>
- Date: Fri, 2 Jun 2000 22:13:20 -0400
- To: "Dan Brickley" <danbri@w3.org>
- Cc: <sw99@w3.org>
There was more to his message than my original forwarding. Tim -----Original Message----- From: Bill dehOra <wdehora@cromwellmedia.co.uk> To: 'Tim Berners-Lee' <timbl@w3.org> Date: Thursday, June 01, 2000 7:49 AM Subject: Longish: RE: RDF/XML/Internet Collisons, Process (was Moving on) >Hi, > >I'll respond thru email, since it may off topic for the uri list. Your post >has re-triggered thoughts of mine from www9. There is no real need to reply >to this (it meanders), I would be sufficiently gratified to know that you >read it. > >The main reason I went to www9 was to ask you what the semantic web was, as >my IChing told me to find a great man. I wanted to do this at the Friday >lunch, but wound up offering a badly phrased answer to someone else's >question (Murray, who gave me the mike was deeply unimpressed: but hey, I'm >a conference newbie, and nobody else seemed to have any answers :). > >As things panned out Eric Miller gave a presentation on lifting thematic >Topic Maps from multiple sites after lunch, at which the long awaited >epiphany occurred (my IChing hedged its bets: it didn't tell me which great >man). So I talked to Eric Miller awhile, and basically told him that the >questions and troubles I had wanted to air at the session. But they seemed >too trivial (I am a coward), in light of the way the open forum had gone. He >left me in no doubt that what was going on with the semantic web was the >right thing to do, and the way forward was in building tools and lashings of >evangelism. > >What I wanted to ask were the following: > >-would it be posible to hang a normative glossary off the RDF(s) docs, in >particular to clarify the following terms: > reification > resource > URI > >so that I can go to that glossary and say 'ah, this is what they mean >by...'. Right now I have to explain to myself that reification is a bit like >the way set families work. If the set S is {{X,Y,Z}, B, C} then C is an >element of S, but X isn't. or maybe it's like using corner quotes in logic >to distinguish between uses and mentions. You see, I don't really trust >myself to understand other peoples meanings properly without some >hand-holding. > >-would it be possible to get a straight forward non-normative explanation of >why RDFs has a flattened type system with one cycle in its definitions (ie >why does it have the same cycle that the Smalltalk-80 object model has). > >-does the XML syntax really have to be like that? What's wrong with just >adopting your simplified syntax? Much of the complexity of RDF is in the >Schema, but the slight obtuseness of the XML in the Model, can easily be >seen as complex and prevent people seeing the elegance of the Model itself. >I remind you that many developers will look at the XML and think that *is* >RDF. Somehow the UML community have not fallen into this hole. No-one >(except CASE engineers) confuses a UML diagram with running code. > > >:I am not sure we don't *need* to sell everyone on RDF. Many >:projects are now getting on board. > >I emphatically disagree with this. If I ever saw a technology that needs to >hit a crystallisation point, that needs positive network feedback effects to >succeed, it's RDF. What RDF can be used for is more important than the *ML >family of languages. It's not at that point yet, and can still fail. RDF >requires ubiquity and democracy as much as HTML ever did. It seems more >important to get it used than get it right. > >WHY RDF MIGHT FAIL - : > >-Assuming it will succeed without much push. 'Hey, it's obvious and >inevitable'. >-Bad public relations work. By this I mean not speaking to the mass of >developers and users of the web, who are not KR/Agent/Logic/AI people. Dan >Brickley said at the semantic web developer day that the RDF standard was >'somewhat schizophrenic'. This is only a bad thing if we insist that one >personality is the real one. >-The 'Not Ontologised Here' (NOH) mindset in the world that shouldn't been >underestimated. >-Not unrelated to NOH is 'I don't need all this hard stuff, why can't I just >use XML?'. This can be a difficult question to answer. > >WHY RDF WILL WIN OUT - BANAL, BUT WHO CARES?: > >-The major search engines have RDF processors, since they are using dumps >from the open directory. You can extrapolate what might happen if they >announced a preference for RDF metadata from some of the bigger sites. >-When/if AOL adopts mozilla XUL as a UI rendering tool, most non-desktop >browsing systems will imitate, and carry RDF data storage with them. >-RDF doesn't require a killer app, just the crystallisation point. >-The AI and Agents communities will accept it, when they realise that loads >and loads of useful and usable information (that's what hits our switches >really, hacking around non-interop is just excruciating) is being stored as >RDF. > > >:I agree RDF needs better explanations and materials. As always it is >:really difficult to knkow how to spend limited resources. > >So, here are my pleadings to you, in light of limited resources: > >-For the next versions of RDF, have three reccs, Model, Serialization, >Schema. in other words formally separate the Model from the XML entirely. >-Have normative glossaries. >-Persuade someone to write an RDF book, or get a publisher to request for >authors. Books tend to make technologies 'real', the way TV makes events >'real'. I bet O' Reilly or Wrox would publish on it. >-Speak more to the masses on RDF/XML. The logicians and librarians aready >get it and are sold for the most part. I'm not sure about Authors and >Developers. Emphasize RDF/XML serialization as a really good way to >catalogues and store and interop 'stuff about stuff'. The cool stuff *will* >follow if there a a bedrock of RDF to mine from. The difference, if you >will, between saying something about something and anything about anything, >is purely a matter of time. > > >I'm selfish really. RDF can help me realise my two pet 'projects'. I hope >someday to build agents that read and speak and think (in RDF) on your >behalf, and can cut the signal to noise ratio of information to human scale >levels. I also hope that someday I can build tools that allow people to find >answers from people the way they find answers from documents today (like a >opt-in Borg). In the meantime I'm happy to write tools to build boring old >catalogues, as my current mission is to make my company 'get it', and small >steps are the best way. But next year I can have some fun with the >catalogues. > > >Yours > >-Bill de hÓra >
Received on Friday, 2 June 2000 22:11:57 UTC