- From: Tim Berners-Lee <timbl@w3.org>
- Date: Wed, 17 May 2000 18:03:27 -0400
- To: "W. E. Perry" <wperry@fiduciary.com>, "Paul Prescod" <paul@prescod.net>, <xml-uri@w3.org>
Thank you, Walter Perry, for defining semantics and distinguishing it from syntax. I am happy with those definitions. I continue to agree with your message up to a point, where my model for communication and yours are I think basically the same but applied differently to this question. I believe that an XML document has semantics. It has meaning. (in general!...) >> * if semantics are entirely local, then does Microsoft have the right >> to interpret the "a" element type in xhtml as meaning "archive" and the >> "b" as meaning "Beethoven"? > >Absolutely, if they can implement desirable behavior from a process by so assuming. I would disagree. I would say that there are a set of understandings which are broken when that happens. Also, the system has broken here. The author of a document uses <b>poison</b> for emphasis and the browser puts up a picture of Beethoven. The reader of the document does not notice the phrase and drinks the poison. Whether you regard that as desirable or not, it was not the semantics of the message. It was not "that signified". That is not the way the Web is designed IMHO. The Internet is designed on the assumption that using common languages such as English we come to define protocol specifications which are generally considered to be well defined. This specifications define new languages (such as xml namespaces) where a syntactically valid sentence has a semantics which is then indirectly defined though the language spec. You find the semantics by applying the definitions from the specification to the particular combination of terms chosen by the document author. The semantics is a form of product of the syntax of the document and the semantics of the specification of the namespace used. In that model the specs are all grounded in English which is assumed a common language. English itself is defined in a complex self-referential way with few actual connections to anything physical. The web might slowly work its way in that direction as more an more semantic links between languages are created but let's not get into that now. But the identity of the namespace is obviously key. So is the concept of "grounded documents" where a grounded document is one in some chosen base set or another document defined in terms of grounded documents. For example, a good web site's web pages are grounded in W3C Recommendations. >> If they write a web browser that archives >> any link you click on and play's music for bold, will you defend them on >> the basis that semantics are local? > >No. If I want that behavior I will buy their software or use their process (or to use Tim >Bray's terminology, dispatch to it from my local node). I prefer to talk about what things mean without discussing behavior. I agree with you that you can read a book, understand well what the writer meant, but behave in a great variety of ways as a result. >> I think that behavior is local, but >> semantics absolutely must be shared. > >Not shared a priori. That is the single great lesson to be drawn from the Internet topology >of autonomous, largely anonymous nodes which, when they act must treat one another as peers >because they do not know enough about each other to infer any other relationship. Semantics >are effectively negotiated in the instance, and the ability of two nodes to negotiate a >successful transaction, understanding, or other disposition of given content on one occasion >implies NOTHING about their likelihood of reaching a similar conclusion, or any conclusion at >all, with analogous content on a subsequent occasion (this really is a Heraclitan cosmos). You rule out the ability to make a promise. But you talk about negotiation. Negotiation normally implies a conversation with a shared state. I don't understand your model of how this works. (In fact there are lots of protocols make promises. When you associate a uuid: with a document you promise never to reuse it. When you order something you promise to pay...) >The vertical industry data vocabularies (FpML, ESteel, etc., etc.) which have been the >shining demonstration of XML's acceptance in the past year are predicated on a closed-world >view which is anathema to the real potential of XML as syntax. All of these vocabularies are >designed to convey intent, with the expectation that intent will be correctly interpreted and >result in the execution of a desired process. I would say that the expectation is often that they are not misinterpreted. Often a document is sent without the knowledge that it will be understood. As you say, to be assured in advance requires a promise. But when you send me a message grounded in documents I am aware of I can infer from it an equivalent message I can understand. > It is only in a closed world (a cartel, to put >it bluntly) that those expected processes could be assumed to be generally known and >generally considered desirable activities. In the Internet topology we simply do not have >enough knowledge of our fellow nodes to make such assumptions, but we may well have the >desire to do business with or otherwise communicate with them. This is surely a statement about social groups not about the Net. There are many groups which share languages apart from cartels. Everyone using a certain proprietary word processor forms such a group. > In order to do so, we could >first attempt to indoctrinate them in the shared assumptions and accepted premises of our own >milieu, but that might not work: they may prove recalcitrant; it may turn out they are more >influential than we within the larger Internet universe; or they may simply not give us their >attention, or understand what we are trying to communicate. That is where the value of the >Semantic Web becomes apparent. One pair at a time, autonomous nodes can build the semantic >context within which they come to understand one another. I will not elaborate on this >process now; I have done it at plenty length elsewhere. huri me an uri? >The only point to make now is that >the negotiation must be based on each node handed the other semantically neutral content. I don't know what "semantically neutral content" is. Messages which do not mean anything? That doesn't help me make sense of your position. [...] Tim BL (apologies for the varying time delays - I am on and offline at Amsterdam)
Received on Thursday, 18 May 2000 02:33:38 UTC