At last, the new layout is ready and deployed. Codename “Dorset2” was completed some days ago but I spent additional time figuring out ways to make a few parts work with Internet Explorer 5.5 and 6. Yes, I know those browsers are obsolete, but IE 6 is the last version that can be installed on Windows NT 4.0, Windows 98 and later (wikipedia) — yes, I know nobody should use anything older than Windows XP for Internet browsing nowadays, hush.

I also had to work around a couple of bugs in Mozilla Firefox 3.5, of all things. Webkit and KHTML-based browsers (Google Chrome understands some KHTML extensions for some odd reason) also displayed some quirks of their own.

Here's a few of screenshots that should display the overall differences between Dorset and Dorset2 (big files ahead!):

• Dorset
• Dorset2
• Dorset2 on Internet Explorer 6 SP1

Naturally, this site is no longer very compatible with IE 5, 6 and 7 because it's using some CSS 2.1 characteristics that are not implemented correctly or at all by those versions. IE 8 works like a charm except for a minor problem with the pre element height rules — which I could fix with a small work-around if I cared enough — but there are also some CSS 3 techniques and/or vendor-specific extensions in use for round borders and text shadows. Nonetheless. I made sure that the site's functionality would not differ between IE 5, 6 and 7, so even if the appearance differs, nothing should work incorrectly.

VirtualBox was very helpful when testing all this stuff. It'd been very hard to run Debian lenny and squeeze at the same time otherwise!

Opera 10.00 showed problems handling multiple children elements with transparent background images. That's a real pity and I hope that newer versions don't have this problem.

The bottom-left corner is not round. There's a good reason for this, and I hope to fix it in the next iteration, some day. For now, Dorset2 is here to brighten and soften your day!

I have recently discovered that the color scheme and overall “look and feel” of an user interface, including web sites, can do a lot with my mood. Two days ago, someone on freenode.net's social channel (#defocus) linked to her blog, which has a black background — that's not bad or unusual, but I noticed that the dark scheme affected my mood making me feel slightly upset for a few minutes. I have no idea if this is just another quirk in my brain's functionality, or normal.

Nevertheless, it does sound like something I could use to my advantage, and to please my somewhat loyal reader (hi Espreon!).

Codename “Dorset2” has been a work in progress since last November. I had experimented with round shapes, box shadows and gradients, using CSS 2.1 and background image tricks, but I didn't get very far due to Dorset's inflexibility and design flaws at the PHP level; basically, I'd have had to edit every single page to adapt them to the new scheme, and that'd be boring and tedious. However, a few days ago, “Poison Ivy” was completed, enabling me to share the basic and simple functional code with three websites, or document sets, so to speak:

• The Wesnoth-UMC-Dev website, (now codename “Kalari”);
• This website (codename “Dorset”); and
• Dorset2, not yet online.

Ivy's design allows me to simply “flip the switch” to convert every web page in my laptop's test Apache instance to use the new scheme, thanks to a extremely primitive, yet effective template and configuration system. A couple of lines of code:

define('DORSET2_ENABLED', TRUE);
define('SKEL_BASE_PATH', DORSET2_ENABLED ? '/dorset2' : '/dorset');

By toggling DORSET2_ENABLED, I can test my code with the old and new templates and stylesheets as necessary, without editing any of the actual pages!

I deployed “Poison Ivy” on the online site last night, so this is already theoretically possible in here... except that the Dorset2 files are not finished or online yet. I did resume my work on it some days ago after finishing Kalari, though.

If a dark scheme can have negative effects on my psyche, what could bright (but not too bright), soft colors and shapes do for me? Basically, Dorset2 aims for a relatively simplistic look, with soft shapes and colors using gradients and round corners for some elements. The color scheme is also slightly brighter than Dorset for some elements; but the shapes are what matters here. A box with round corners and no solid border makes the contents look soft to me; compare current Dorset which uses (way too many) rectangular boxes with solid and dark borders everywhere, inside and outside the main body.

Here's a (rather big) screenshot of Dorset2. Apologies for the admittedly awful rendering of Verdana Bold; that must be freetype's fault.

• Dorset2 (PNG screenshot)

Since it's a work in progress, I have not gotten around to tweaking the CSS to make it work as best as it's possible with Internet Explorer. It doesn't look too bad at first glance, but it gets worse at the bottom (not pictured) thanks a gradient background trick that makes some text disappear at random in IE 6 SP 1 — and for whatever reason, this doesn't affect IE 5.5 or IE 6 SP 2 and later. I figured that I'll make my work easier for now if I write rules to disable certain decoration elements with these broken browsers.

Hopefully this gets finished soon.

Long, long ago, I talked about several issues I had with Mozilla Failfox Firefox 3.0 and openSUSE 10.3 for the AMD64/EM64T architecture.

Ever since then, I have learned several things:

1. Debian's Iceweasel fork doesn't seem to be much ahead of mainline Firefox in terms of bugfixes, as far as I can see. This might be not true for security fixes and such; I admit I haven't done any actual research on this and I'm basing this statement on my user experience.
2. The Download Day was a trap.
3. Other people who I have talked to regarding Firefox's stability on Linux claim that is never/rarely crashes, but all of them use x86 kernels and userspace.
4. Iceweasel 3.0 taints the Debian GNU/Linux 5.0 Lenny distribution on the AMD64/EM64T architecture, with no differences in either of my laptops. This Linux distribution is remarkably stable otherwise, and lived up to my expectatives since I originally switched to it when it was the Testing distribution — this is, comparing it to the released openSUSE 10.3.
5. Off-line browsing is truly, horribly underestimated, to the point that one of the major web browsers does not support it at all; probably in favor of simplicity and ease of use, and “permanently connected people”. But, STILL...
6. It's not a good idea to leave a chainsaw and a newspaper near the reach of a cow.

I recently switched to Debian Squeeze, which is still under development (e.g. Testing) as of this writing. Originally, I just got a newer revision of Iceweasel 3.0 with the set. Some weeks ago, I got upgraded to 3.5.

As I mentioned in my previous post in this series, the status bar does still glitch a lot — no, wait — the status bar glitches even more than in 3.0. Scrolling is less laggy but only with smooth scrolling disabled, although I am not exactly using a well-supported video configuration at the moment and I probably should not complain about performance issues with any 2D application unless I'm willing to use the unaccelerated X.org VESA driver for benchmarking or shut up.

The Live Bookmarks feature stopped working after the upgrade until I went and manually reloaded every single Atom/RSS feed I had linked in a neat folder in the bookmarks toolbar. It took me a while to realize that nobody posting anything near Christmas was a bad sign — I didn't miss much anyway, since my feed sources aren't really chatty. Yes, I know I'd be better using an actual feeds reader, but I'm just that lazy, which is also why I don't use Opera as much as I want.

However, this version of Firefox is much, much more stable than 3.0 — as far as this AMD64/EM64T architecture user is concerned, that is. Firefox just got better, really. But it's still rather odd because I've heard comments on IRC of people claiming that it got more unstable instead. Hmm... Well, maybe Windows or x86 Linux users are less lucky this time?

Firefox 3.5 also supports the CSS text-shadow property, which was introduced in the CSS level 2 specification, removed in revision 1 (CSS 2.1), and seems to have been picked up again for CSS 3. No version of Internet Explorer before and including 8.0 supports this (although ISTR that they support a shadow filter using a custom extension to CSS that didn't even follow the specification for naming vendor-specific properties), and current Opera, Safari and Chrome support this property well. That means that I must make more use of it in this site's stylesheets from now on.