- From: Preston L. Bannister <preston@bannister.us>
- Date: Sat, 28 Apr 2007 21:08:33 -0700
- To: "Matthew Ratzloff" <matt@builtfromsource.com>
- Cc: public-html@w3.org
- Message-ID: <7e91ba7e0704282108m10c32aa8wca3fb8f14f782678@mail.gmail.com>
On 4/28/07, Matthew Ratzloff <matt@builtfromsource.com> wrote: > > > > On 4/28/07, Preston L. Bannister wrote: > > Note that as a developer writing applications for the web, I very > strongly > > agree with the hard line. Versioning to opt-in to changes in behavior > is > > essential. This judgment comes from the experience gained from years > > (decades) of shipping multiple versions of multiple applications, and > from > > insight gained in talking to customers. > > > > What works in practice is different from what might work in theory. > > This is one of those lessons learned from experience ... and requires > > a bit of not-entirely-obvious insight. > > > > So the Microsoft folk are not alone in this judgment. > > Instead of citing your years of experience (which are relevant, but not an > argument in and of itself), please counter the argument for a "bug mode" > with actual examples and points of disagreement. I agree that experience is considerably less than proof. :) Can't say I have made any attempt to record experiences in detail. Can say when things have gone wrong I've thought carefully about what went wrong, and how to preclude making the same mistake in the future - and that this has paid off (many times). Hence my expectation that the IE-folk (if they were paying attention) would come to the same conclusion. Might be mistaken, but I think Chris Wilson was offering an example from his experience with IE7 (at about the time I had to drop off the teleconference for another call). The most relevant examples to this domain are going to come from the IE folk, due to their longer history and larger market. Please do not discount their observations, just 'cause they are Microsoft folk. :) If you absolutely want to link your product to the then-current browser > version, why not pre-emptively include: > > <meta http-equiv="Content-Rendering" content="IE 8.0" /> > > ...or serve the equivalent HTTP header? We don't need this. When IE comes out with support for HTML 5, the mass of new web pages written to HTML 5 will mainly be tested in that version of IE. In effect pages declaring: <!DOCTYPE html PUBLIC"-//W3C//DTD HTML 5.0//EN"> (or whatever equivalent we decide on) will come to mean HTML 5 as implemented in that future version of IE. If we do our job as the HTML WG well, that version of IE (and other browser implementations) will only differ from the standard only in areas not well covered by validation tests. If we are conservative in adding new function, and thorough in requiring validation tests for any new function, then the chance of uncaught differences becomes small (and even smaller in future iterations). BTW, using the label "bugs mode" offers the wrong connotations. Better to think of this more as "make existing pages work" mode.
Received on Sunday, 29 April 2007 04:08:36 UTC