- From: Jacek Kopecky <jacek@systinet.com>
- Date: Thu, 13 Dec 2001 14:06:35 +0100 (CET)
- To: Andrew Layman <andrewl@microsoft.com>
- cc: xml-dist-app <xml-dist-app@w3.org>
Andrew, I think you hit the point exactly with your analysis. My opinion, in terms of your analysis, can be summarized as the following: I believe in 1a. I'm willing to bend a bit by actually allowing external references (like to attachments etc), the graph still being closed and deserialized in its wholeness. Because I'd allow external references, I see the typing information for the referred data as being provided by the resolving process in an implementation-specific or third-party-specified way (for example a .net impl could guess depending on the filename extension, and SwA could specify typing according to MIME types of the attachments). Unfortunately, this 1a or 1b/2 view of SOAP Data Model and Encoding is a majority vote thing, I guess, not likely to end with consensus. Best regards, Jacek Kopecky Senior Architect, Systinet (formerly Idoox) http://www.systinet.com/ On Wed, 12 Dec 2001, Andrew Layman wrote: > I've just read all of this thread again. > > There are several issues raised in it. > > One is the question of whether one can somehow, by mere inspection of > the URI value of an href attribute, determine the type or other > characteristics of the resource that might be obtained by resolving the > URI. As Henrik described, the answer to this is "no." One might > determine the type, etc., if one had additional information, say > provided by a schema somehow or by a yet-to-be invented attribute, but > no such mechanism is now part of either the URI or the SOAP encoding by > itself. > > I think that a larger issue is the apparent clash between two different > assumptions about what is in a SOAP message within an element containing > content labeled with a soap-enc:encodingStyle attribute: > > 1. It is an abstract graph of data, and it must be reconstituted by > a reader as exactly that graph of data. > 1a. The graph is closed. Hrefs are only a serialization > mechanism for encoding internal links. External links are data as far > as the serialization algorithm is concerned, and hence never appear as > href values (or, if they do, this is peculiar and should be specially > distinguished from the internal links). > 1b. The graph is open and infinite, and any serialization is > only a fragment of the cosmic, infinite graph. > > 2. It is an XML structure, which the writer advises may be turned > into a graph by a defined process. The graph is open, and the > deserialization process will result in some links being immediately > resolvable to nodes in the graph while others may not be immediately > resolvable. > > I believe that those who think mainly in terms of RPC hold 1a, that RDF > holds 1b and that those who think mainly in terms of XML documents would > hold number 2. > > 1a leads naturally to the idea that external and internal links should > be syntactically distinguished, that internal links MUST be resolved by > any reader (in fact, that actual deserialization MUST be performed by > any reader) and the failures to successfully resolve an internal link > indicate a gross error somewhere. > > I think that 1b and 2 amount to very similar if not the same processing > in practice, namely that if the reader finds that deserialization is a > convenient way to process data, he may deserialize the XML into a graph > by a defined process, and will end up with some references that are > immediately resolvable and others for which attempts to resolve are > deferred. What actions a reader takes respecting these depends on the > semantics of the data and the purposes of the reader. > > I hope this analysis is helpful. > >
Received on Thursday, 13 December 2001 08:06:38 UTC