- From: Martin Duerst <duerst@w3.org>
- Date: Sat, 23 Feb 2002 04:08:20 +0900
- To: www-forms-editor@w3.org
- Cc: w3c-i18n-ig@w3.org
Dear XForms WG, I found some more time to look through XForms. Below are the most important comments. I have a lot of small editorial comments, too, but I suggest that it's more efficient if I sit together with some of you next week in Cannes. I have copied the I18N IG because many of my comments are I18N-related; the I18N WG may want (or not) to endorse some of these comments, although that will not be possible before the end of the official last call period. [I18N IG: Please note that the www-forms-editor@w3.org is publicly archived.] - IRI/anyURI: There should be a note that data of type anyURI, and xlink:href, are internationalized. Otherwise, it's easy for implementers to miss this. A wording like in XML Signature (sorry no pointer, currently offline) is a good start. It could go into 3.2 and/or 3.8. The spec should make clear that the conversion (->UTF-8->%HH) only happens when anyURIs get resolved, not in XForms or as part of an instance. - Bidi: I remember Jonathan Rosenne once making a comment about a problem with bidirectionality and HTML forms. I think it was that the base direction of an input field, as used by the user, could not be transmitted to the server. This is also a problem for XForms. Even more, there is no way to set the base directionality for an input field. This looks like a step back from HTML. (Please note that the base directionality of a text is related to the semantics of that text, and not just a presentation issue.) [A generalization of this is complex types as single form controls, e.g. for HTML fragment input,... My personal opinion is that the WG should at least have an idea of how they plan to add that in a later version.] - In 2.2, an example with a Date is given. The HTML example is not very good, because usually, there are separate fields for month and year of the expiration date. An XForms example of the presentation to the user is not given. If the user is supposed to type in 2002-03, this just won't work anywhere in the world. It is also not very clear how the gYearMonth type would get translated into a reasonable UI component (in particular because it may be localized, where e.g. in Japan, the year would come first, with the consequences that many people would interpret the first two digits as the year rather than the month. There are several more problems with dates. In particular, I think that in section 8.1, it is very important to say clearly that a simple text field is not a good idea for date input. - 7.4.1.1: Only allowing 'true' and 'false' for boolean values is more language-dependent than necessary. '0' and '1' should also be allowed, also to fit with XML Schema. Regards, Martin.
Received on Friday, 22 February 2002 14:11:40 UTC