- From: Timur Mehrvarz <timur.mehrvarz@web.de>
- Date: Tue, 9 May 2006 13:13:53 +0200
- To: Anne van Kesteren <annevk@opera.com>
- Cc: "Kevin E Kelly" <kekelly@us.ibm.com>, public-cdf@w3.org, member-cdf@w3.org
Hi Anne. On 3. May 2006, at 20:40, Anne van Kesteren wrote: > * When a new version of XHTML is issued, say XHTML 1.2, we would > have to update the URI to some new string. I don't think we *have* to do this. While today, the WICD profile may look like a maximal feature set, this will not stay this way and it doesn't have to. The WICD profile would ideally even evolve into a least common denominator feature set. Different agents will support more features and newer specifications. > UAs can't just remove the old string as content is still checking > for that string. That's a serious issue. Each time a new version of > something is introduced and a new string is issued it has to stay > permanently on the UA its Accept: header because otherwise content > that isn't updated (lots of it isn't) will break. Also given that > we currently only cover XHTML, SVG and CSS and likely want this to > grow there will be lots of updates over time making it not just a > 64 characters growth, but 64*n characters growth where n is a > pretty big number. The problem is here that current content can't > be future proofed. They don't know the string might change and > don't know how it might look otherwise. So my first argument > against it would that the proposal isn't forward compatible in any > way. We need to educate authors, to only check for the "/wicd" string in the profile URI and (if really required, for content depending on features of a next version) to also check for a minimum date signature, like for "/2007/03/" instead of "/2005/12/" (doing: year>minY || year==minY && month>=minM). > * As pointed out, that's not only a problem for the UA, that's also > an issue for content. I can currently check for WICD Full 1.0, but > at the moment SVGT 1.3 is released or XHTML 1.2 and eithers makes > some backwards incompatible changes I would like to know that as an > author, but I can't really... As an author, you would like to know. OK. Using the profile date mechanics, described above, we could create as many profile update specifications as necessary. > * It's unclear to me when UAs can add this to their Accept: header. > For example, no UA supports the longdesc="" attribute on the <img> > element in XHTML, yet we expect XHTML to be fully supported because > otherwise the UA would not issue use this string, right? There are > many tiny things that are not fully implemented which will probably > not be fixed anytime soon, are we going to delay WICD until 2010 so > we can be sure they are? Yes, we have to talk about this. One possibility is simply trusting vendors. Another one is for vendors to pass a testsuite. A testsuite will not test everything, of course. It would probably not check for img/longdesc. (2010 would be a little late.) Timur
Received on Tuesday, 9 May 2006 11:14:35 UTC