- From: Dr. Olaf Hoffmann <Dr.O.Hoffmann@gmx.de>
- Date: Tue, 4 May 2010 11:35:21 +0200
- To: www-svg@w3.org
Patrick Dengler: >We are working on coming up with clear strategies for Internet Explorer and >hope to update developers on that soon. I would, however, suggest that you >avoid UA string detection and leverage the built in hasFeature in the DOM to >detect SVG support, as well as more granular hasFeature in SVGDOM, and >declarative <switch>. What about the HTTP_ACCEPT string to get a rough impression about an overall interpretation of SVG? (can be problematic as well, if not all user-agents with SVG capabilities or active SVG plugins indicate this within this string ...) The problem with switch and requiredFeatures is often, that the feature strings are not precise enough. Therefore and often for other reasons the reaction of the viewer is not precise enough. Cautious viewers will too often use the alternative. Careless viewers will too often claim to interprete something correctly. Another problem is, that older viewers do not know newer feature strings. In many cases their presentation would be nevertheless acceptable. Theoretically requiredFeatures is a nice concept, practically it is hard to use with the current diversity of viewers and their individual behaviour and interpretation of this feature. For example should currently really any viewer switch for 'http://www.w3.org/TR/SVG11/feature#Animation' properly? - all have bugs and gaps concerning animation, therefore a conditional processing should be never true theoretically. On the other hand, for a more or less large subset the interpretation in several viewers meanwhile is correct and usable (Opera, Bitflash, Batik/Squiggle, WebKit, Adobe plugin, current Firefox preview etc). Therefore practially the processing of this string is not very informative. The behaviour for requiredExtensions is even harder to predict. Currently I did some tests with foreignObject. And it turned out for example with a simple poem marked up with DAISY and styled with CSS, it is better not to use requiredExtensions for Geckos, just because they manage to display the styled text in an acceptable way. Opera on the other hand does not provide a useful result. This should apply for arbitrary XML styled with CSS as well. For XHTML with some functionality in foreignObject Opera tries to present something, but it is useless - the alternative would have been better. Presentation with a Gecko is much better, but the function does not work, again to present the alternative would have been better. With WebKit in Arora neither the XHTML nor the SVG alternative is presented. With some simple MathML presentation in Geckos was nice, with Opera questionable. Opera should have presented the alternative for a better result. With WebKit in Arora neither the MathML nor the SVG alternative is presented. Just from these few tests I got the impression, that an average author will be overstrained by switch and requiredExtensions. The new requiredFormats has problems in some viewers as well. For example in Opera one has the problem, that this with a value like 'application/ecmascript' does not switch to a scriptless alternative, if the user only switched off script interpretation. And older viewers not knowing the attribute of course ignore it anyway. Similar applies for the related feature string for scripting. Because some viewers can have different methods to present SVG content (for example Konqueror 4 can use the build in KSVG, WebKit or a plugin like that from adobe), the information within the user-agent-string indeed is not very useful. Additionally due to some bad history it is often faked to indicate Netscape 4 or an MSIE that claims to be Netscape 4 ;o) However due to the problem of previous MSIE versions with XHTML many authors are already used to sniff for problematic user-agents. Hopefully many of these algorithms are optimistic for newer versions and unkown user-agents not faking other user-agents... (and it could be a good idea of course to not longer claim that a newer MSIE version is only compatible with Netscape 4 with almost no CSS capabilities and no SVG capabilities ;o) Therefore currently I can see no simple strategy for authors to work around gaps and bugs in different viewers. The only more or less useful approach is to provide a text alternative with title and desc elements. With this at least the audience has an indication for the quality of the presentation of the currently used viewer. Olaf
Received on Tuesday, 4 May 2010 09:40:08 UTC