- From: Marcos Caceres <marcosc@opera.com>
- Date: Mon, 27 Sep 2010 13:41:13 +0200
- To: Doug Turner <dougt@dougt.org>
- CC: public-geolocation <public-geolocation@w3.org>
On 9/27/10 7:59 AM, Doug Turner wrote: > It is clearly under-specified. > > Do you think that there is an expectation that the exceptions need to > match for browser interop? Because current implementations are all throwing an exception, then in terms of interop, the risks of changing the spec are minimal (look at this as a clarification). It only gets tricky if the developer wants to check why the exception was thrown - in which case, yes, there are (minor) interop issues as demonstrated by my last email. So, at the very least, the spec should say that an exception is thrown. The only way to understand that this is what is intended by the spec is to turn to the test suite: the test suite is acting as the normative text for the spec - which is bad. If an exception is thrown, then the type of exception should be specified. Other specs clearly define what the exception to be thrown should be in given situations (see any of the "HTML5" related specs). We have a strict QA system that requires us to specify not only that an exception is thrown, but also check that the right kind of exception is thrown. That is the reason I raised this. Kind regards, Marcos -- Marcos Caceres Opera Software
Received on Monday, 27 September 2010 11:41:52 UTC