- From: Martin Duerst <duerst@w3.org>
- Date: Wed, 21 Jul 2004 11:04:50 +0900
- To: www-voice@w3.org
- Cc: w3c-i18n-ig@w3.org
Dear Voice Browser WG, Below please receive the I18N review of CCXML from the I18N WG. We are sorry that this review is late. Please send all replies not only to myself, but to the I18N IG mailing list (w3c-i18n-ig@w3.org). [1] In all places where the spec uses URIs, IRIs should be used. The Schema should already define this by the use of the anyURI type, but this has to be pointed out in the text. [2] There are some provisions for creating a log. In what character encoding will this log be written? We suggest to specify an Unicode- based encoding such as UTF-8 to avoid encoding propblems. (CCXML could log any character, but if an encoding such as iso-8859-1 is used, some characters cannot be written out, assuming the log file doesn't have an escaping syntax) [3] How is the language of error messages (reason) determined? for example in 6.3.3, 6.3.6, and other places. Different languages may be required for communication with the user (that could be defined as using the value of xml:lang inherited at this point in the document) and for writing to a log (that might desirably be in a single language even if the server serves users in many different languages). [4] The xml:lang attribute has to be available on any element that can (directly or indirectly) contain natural language text. [It should also be made clear that this language info doesn't apply to documents pointed to from a CCXML document.] [5] Section 6.2.2: <meta> This section should very clearly state that for CCXML, the 'encoding' pseudo-attribute in the XML declaration is the only way to indicated the character encoding of a document in the document itself, and that something like <meta http-equiv='Content-Type' content='application/ccxml+xml;charset=foo' /> is prohibited and to be ignored by the client. This also applies to http-equiv='Content-Language'. [6] Section 6.2.6.2 <createccxml> [this comment applies to all other cases of URI construction] The 'namelist' indicates a list of variables that is sent as name/value pairs to the server. How exactly is this done? For example, what is the separation character between name/value pairs (it should be ';', not '&'). More specifically, how are variable names that contain non-ASCII characters transferred to the server? How are variable values that contain non-ASCII characters tranferred to the server? We very strongly suggest that the spec make the use of UTF-8 (plus %-encoding for URIs) mandatory, independent of the encoding of the document, otherwise, this will be a mess. The other details of the encoding of name/value pairs into URIs should also be specified carefully and in full detail. [7] 6.2.7.2 (and maybe other occasions): Avoid colloquial expressions such as 'oops!', which make the spec potentially difficult to understand for people not of English mother tongue, and difficult to translate. (potential other examples: file name 'gimee'). [8] section 8.2.2: <script> after the example using a CDATA marked section, please make clear that other means of escaping, if necessary at all, are also possible. (in the example, the two occurrences of '<' could be replaced by < ???? is it necessary to also replace '>' by >????). Or better even, change the example to use these escapes. [9] Section 8.2.2.2, charset attribute: It should be made clear that any charset attribute given in the HTTP header, or more generally, any external charset information, and after that any charset information in the retrieved document, if the document is able to contain and contains such information, has priority. [10] 9.2.3.2, data attribute: Why is this limited to US-ASCII? We think that this may be an unappropriate restriction. [11] section 11: there should be examples that show the international capabilities of XML and CCXML [12] vm.vxml: 'voice mail hell' might not be appropriate in some cultural contexts Regards, Martin.
Received on Wednesday, 21 July 2004 02:15:07 UTC