Christian Heilmann wrote: > Make the sites work in all browsers, enhance by object detection or > valid CSS selectors and do the cleaning up for those your users are > most likely to have according to your stats.
I agree with the principle Christian outlines here, but I think the danger in the wording used here and in other responses is the definition of the word "work" assumed by the reader. If "work" is taken to mean looking the same more or less to the pixel, then this is not possible given the browsers listed in the OP's stats. Netscape 2? Even table layouts won't work reliably, I suspect. I believe that most professionals would define "work" as in providing access to the content and basic functionality, while reserving visual design consistency for those browsers capable of displaying it. In practice, this means using well-structured valid HTML, with CSS delivered selectively to those browsers capable of using it. This should not require any scripting on the client or server side, unless one wants it to. I believe a common approach is to support the version 5 browsers and up visually, to the greatest extent you can. Personally, I support the aim of making the most common use cases work well (IE6 Windows, usually), but not at the expense of the better browsers. Using media attributes (e.g. media="screen") in the link tag or importing styles is a good way to hide all the CSS from NN4 which will often make a page unusable if it can see the styles intended for other browsers. So to summarise; IMHO version 5 browsers and up should get as close to the same visual experience as one can get it, subject to available resources and time available, while earlier browsers should receive unstyled HTML. There is no reason at all why one should not use simple layout tables for the rapid development and delivery of a site, if one lacks the experience to build the same design in CSS in a reasonable timescale. In reality, only the web designer will know whether the site is CSS layout or not. No user will know or care. @ Dova - You mentioned using tables and DIVs together; in fact, in this hybrid style of web design (cf. Zeldman [1]) the TDs replace DIVs in most cases; you can use the same class and ID selectors as you would in a CSS layout, and enjoy the best of both worlds while you work on an all-CSS version to hone your CSS skills. The unknown browsers in your stats are almost certainly search engine spiders and other robots. Using JavaScript-based stats like Shawn Inman's Mint avoids these gross distortions, while introducing other, more minor ones. It also looks like there are other irregularities - NN2 should not be there; NN4 is no way more common than IE5.5, and so on. Or - either you have a very small, unrepresentative sampling of the general browser population, or folks have some strange browsing habits down there in Virginia. :) Cheers Ian [1] Designing with Web Standards, Jeffrey Zeldman 2003, New Riders Chapter 7, page 181 -- _________________________________________________ zStudio - Web development and accessibility http://zStudio.co.uk Snippetz.net - Online code library File, manage and re-use your code snippets & links http://snippetz.net ______________________________________________________________________ css-discuss [EMAIL PROTECTED] http://www.css-discuss.org/mailman/listinfo/css-d List wiki/FAQ -- http://css-discuss.incutio.com/ Supported by evolt.org -- http://www.evolt.org/help_support_evolt/