- From: Ian Hickson <ian@hixie.ch>
- Date: Thu, 26 Jun 2014 16:18:52 +0000 (UTC)
- To: Glenn Adams <glenn@skynav.com>
- cc: Boris Zbarsky <bzbarsky@mit.edu>, WebApps WG <public-webapps@w3.org>
On Wed, 25 Jun 2014, Glenn Adams wrote: > On Tue, Jun 24, 2014 at 8:28 PM, Ian Hickson <ian@hixie.ch> wrote: > > > > Compraing implementations to anything but the very latest draft is not > > only a waste of time, it's actively harmful to interoperability. At no > > point should any implementor even remotely consider making a change > > from implementing what is currently specified to what was previously > > specified, that would literally be going backwards. > > That sounds reasonable, but its not always true (an exception to every > rule, eh). For example, in order to ship a device that must satisfy > compliance testing to be certified, e.g., to be granted a branding > label, to satisfy a government mandate, etc., it may be necessary to > implement and ship support for an earlier version. For pointless certification purposes, you can use any random revision of the spec -- just say what the revision number is and use that (and honestly, who cares how well you implement that version -- it's not like the testing process is going to be thorough). Don't ship that, though. Whatever you ship should be regularly kept up to date with changes to the spec as they occur. (It's not an option to not be able to ship fixes, since otherwise you'd be unable to fix security vulnerabilities either, which is obviously a non-starter.) What you ship, and subsequent revisions thereto, is what you should be spending any serious amount of time testing. And for that, you shouldn't use a snapshot, you should use the latest revision of the spec. For the pointless certification, just as for the patent coverage, we should publish whatever revision we have and just stamp it as a REC. It doesn't matter what bugs it has. We know it'll have bugs -- the day after it's published, maybe even earlier, we'll find new bugs that will need fixing. It doesn't really matter, since it's not for use by implementors, just by lawyers and pointless certification teams. -- Ian Hickson U+1047E )\._.,--....,'``. fL http://ln.hixie.ch/ U+263A /, _.. \ _\ ;`._ ,. Things that are impossible just take longer. `._.-(,_..'--(,_..'`-.;.'
Received on Thursday, 26 June 2014 16:19:15 UTC