- From: Xavier Franc <xfranc@online.fr>
- Date: Mon, 02 Feb 2004 14:43:30 +0100
- To: public-qt-comments@w3.org
well, I rest my case, since the situation is considered acceptable. The bottom line is yet another implementation dependent behavior and interoperability issue -- on doc() and collection(). But what can I do ? -- except keep the non-standard catch-error function in my engine! I find a little strange to delay an essential feature like error catching to a distant future for the reason that its interaction with transactions is not clearly seen. Is it such a new and unexplored problem ? Following the same mode of reasoning, Java would have appeared without try & catch ! (because transactions in Java are just an additional package introduced in version 1.4) In addition, it is very difficult to add a feature like this in a language after years of utilization. Best regards to all of you. Sorry to play the role of the grumbler. >We did look very hard at this issue. We decided that it would be a >mistake to tackle error recovery before defining the update facility in >XQuery, which is likely to require introducing concepts such as >transactions and rollback. We do know that it's a gap; you can't do >everything in version 1.0. > >In fact you can't do everything in version 2.0 either; we've tried to >tackle this in XSLT and haven't come up with a design that everyone >finds acceptable. > >We took a look at all the dynamic errors that can be raised, to identify >those where there is no preventative action available to the programmer. >We found two such cases: errors in casting (e.g. converting a string to >a date), which we made preventable by adding the "castable" operator, >and errors in retrieving external resources using functions such as >doc(). The specification for doc() allows implementations a great deal >of freedom, and many implemementations are likely to provide mechanisms >such as the JAXP URIResolver hook to allow users to tailor the behavior. >A URIResolver can be used to recover from errors in retrieving a >document, for example the URIResolver can return a standard error >document containing information about the error. > >We realize that a real try/catch, integrated with transactions and >rollback, is needed in the longer term, but we feel that the current >situation is acceptable for XQuery 1.0. > > >
Received on Monday, 2 February 2004 08:45:54 UTC