- From: Jacek Kopecky <jacek.kopecky@deri.at>
- Date: Fri, 21 May 2004 10:45:01 +0200
- To: Mark Nottingham <mark.nottingham@bea.com>
- Cc: www-tag@w3.org
Mark, members of the TAG, (not cross-posting to xml-dist-app) this is a fine example of a TAG issue. 8-) I've always been of the opinion, and I voiced it while I was still a member of the XMLP WG, that while formally it's cleaner to do what the XMLP WG has decided to do, for the longer term it would be better to disconnect the layers of SOAP, Infoset and XML. AFAIK Infoset has not been updated to point to XML 1.1 and my understanding of it (I haven't made a word-to-word analysis) is that it supports both XML 1.0 and 1.1 serializations, even if these could have different actual capabilities. Therefore if SOAP only says it uses Infoset and if the binding said, admittedly somewhat vaguely, that it serializes the Infoset as XML (pointing both to XML 1.0 and to XML 1.1, the latter link being added as an erratum to SOAP 1.2 Rec), few actual problems would occur. Yes, an XML 1.0 stack would not be able to accept an XML 1.1 envelope. In the not-quite-so-long term, the market forces would either drive XML 1.1 to obscurity or to major adoption, either way, the situation would heal itself. On the other hand now the wound is being made explicit by the XMLP WG and so even when XML 1.1 does, in fact, become widely adopted, SOAP 1.2 stacks will be forbidden from using XML 1.1 features, unless another explicit change (either next version of SOAP or an erratum reversing this one) is made. Same suggestion from me to XML Schema - an erratum saying that it's up to the consumers of the schema what version of XML they use, applying the appropriate feature-sets and restrictions stemming from that choice. To conclude, the higher-level specs should use XML as an actual independent layer, the interface being the Infoset using the highest-available version of UNICODE. Then the choice of using SOAP (or XML Schema or whatever) will not hinder the choice in the underlying layer of XML serialization. Best regards, Jacek Kopecky Digital Enterprise Research Institute http://www.deri.at/
Received on Friday, 21 May 2004 04:48:43 UTC