- From: Sanjiva Weerawarana <sanjiva@watson.ibm.com>
- Date: Sun, 14 Apr 2002 06:52:14 -0400
- To: <www-ws-desc@w3.org>
"Bob Cunnings" <cunnings@whitemesa.com> writes: > FWIW this is one of the things that I think WSDL 1.1 did right... with all references QNames, there is no chance for ambiguity, esp. when importing constructs from foreign namespaces. It's simpler to implement, and I'm not sure that it's any less convenient really... I think it makes complex docs more intelligible to a human reader if all names are qnames. > I also like the current approach, but Keith Ballinger brought this up as a big interop issue. Apparently many people interpret this incorrectly. So one possibility is to just improve the writing quite a bit to make it completely clear that QNames are required. I have also gotten complaints saying that this is not the way references are done in most languages .. people interpret the NS URI as being like a "type" for the thing being referred to and the local part as the variable name. In other languages you indicate the full thing only at declaration time and not at reference time. I don't agree with this logic, but not this is the argument I've heard. Sanjiva.
Received on Sunday, 14 April 2002 06:55:11 UTC