Re: I'm back.

[Joakim Ziegler <joakim ximian com>]

On Sat, 2002-03-02 at 00:05, Jeff Waugh wrote:
> <quote who="Joakim Ziegler">

>> Total number of hits yesterday: 1'612'859

>> Total number of hits from browsers known to handle CSS1/CSS2 in a sane
>> way: 

>> Mozilla/5: 845'834
>> MSIE 5.0:  167'264 (this is pushing the boundary of "sane CSS", but it's
>> close enough)
>> MSIE 5.5:  149'064
>> MSIE 6:    252'491

> Wow, that's really cool. It's remarkably different to the situation a while
> ago, when we had similar results but with IE in the (very front-running)
> lead.

That is cool (although I don't care about IE vs. Mozilla so much, since
the more Windows people who look at our site, the more we will convert).

The point here is that if you put a little care into the stylesheets,
all of these browsers will work well with the same stylesheet. The odd
one out is MSIE 5.0, but from 5.5 and up, the same stylesheet, even with
lots of box positioning and stuff, renders the same in IE and Moz,
almost to the pixel. There are microscopic glitches in IE 5.5, but in
6.0, I've been unable to see any differences that are not caused by the
typefaces.


>> I think it's a mistake to assume that this has to look exactly the same as
>> the CSS version, it'll only serve to cripple both versions.

> I *strongly* agree, and have been saying this for aaaages now. :) I also see
> it as a positive for maintenance, because instead of having to deal with
> many little breakages and limitations trying to keep all of the browsers
> sane (even when you're just dealing with the CSS), you can deal with what's
> best for the browser head on.

Also, with the number of users on non-CSS browsers declining, we can let
the non-CSS version of the page slowly become a sort of lynx/links
optimized version, like the light version many sites have. It'll be a
good option for people browsing on slow links too, for instance people
in third-world countries, who I think it's important to reach.


>> So in view of all this, I think it would be wise to use browser detection
>> to choose between two layouts, one that uses CSS1/2, and is tested on IE5
>> and up (with a focus on IE 5.5 and up), and Mozilla, and one that's
>> tables-based and simpler, and works in HTML 3.2 browsers.

> Currently, with Steve's HTML/CSS combo, we can just switch CSS based on the
> browser, which is attractive to me. I'm trying as hard as possible to aim
> for the ability to serve static html, and very little dynamic stuff.

This uses Javascript currently, right? That strikes me as a little
unelegant. However, see below, I think we can fix all this in one blow.


> Having to serve out different HTML means that *every* page has to be
> dynamic. But that's okay, and if it solves the design issues, I'm happy to
> work with it (it does make some server-side stuff simpler anyway).

I read this, and I started thinking, "Isn't there a way to do this
easily, using PHP or something seems like vast overkill". And then, I
remembered: There is. mod_rewrite. mod_rewrite can rewrite URLs based on
a huge number of criteria, one of which is regular expressions on
arbitrary HTTP headers, which includes HTTP_USER_AGENT. So there's your
answer. Easy, really. We can add one rewrite rule for the HTML files,
which rewrites, say, index.html to index.css.html, and another for CSS
files, which rewrites stylesheet.css to stylesheet.moz.css (or
whatever). If we do this, we get everything fixed automagically, and
there's no need to actually run scripts or use Javascript or anything
like that. Now, mod_rewrite is a bit of voodoo, but not more than I
should be able to handle if I sit down and tweak it a bit.


> What you've suggested will ring in Steve's ears with pain, because I've been
> trying to convince him to do this for a long time. I think it's the best
> solution, because it means we only have to generate 2 * 'the number of
> language translations' pages.

> [ The build system goes through these steps:

>   1) check out current cvs

>   2) build to a temporary directory

>     a) build HTML 3.2 documents in each translation

>     b) build CSS-happy documents in each translation

>   3) move old root directory out of the way

>   4) move new build directory to root location

>   5) delete the old directory

> That way, we won't have the same situation with old junk files that we have
> now. ]

Out of curiosity, how do translations work now? Remembering the gettext
discussion we had the last time that came up, I mean.

The process looks fine, though. So the source files are XML? A custom
sort of DTD? How will the navigational structure be defined? I know
you're planning to present this at GUADEC, but it'd be great to know
more about this before that, for those of us who actually are going to
work with it. :)


>> They've been prettified a bit since then, though, I like the subtle
>> footstep background and the highlighting, although I'm less certain about
>> the use of icons in the sidebar. They're cute, but they don't seem to add
>> much value, and the folder/document metaphor is pretty broken for a
>> website, because a category usually has both a top-level index page, and
>> subpages, unlike a folder, which doesn't have any information itself, just
>> sub-contents. Also, the implication that you're actually *writing* to the
>> page that's currently active is pretty broken.

> That sidebar will not be a tree in the final version - it just looks perky
> on Steve's pages at the moment. ;) This makes of the things you've mentioned
> in this paragraph, and the ones I've snipped, less relevant.

Oh, it won't? Why not? It seems to me that a tree would probably be a
good idea, that's how most sites do it. At least if the alternative is
to only show the siblings of the current page.


>> "Welcome!" in the top nav bar is a little weird. I assume it goes to the
>> front page? If so, it should probably be called "Front Page" or "Main
>> Page", although we might consider dropping it completely, given the common
>> use of the logo as a front page link, as discussed above.  Dropping it
>> will help with page width considerations.

> I'm not sure the sidebar and heading bar are being described will in Steve's
> mockups - I don't really think they'd be used for quite the same things.
> More suggestions on this, and information architecture (I really have to
> send my current summary of this to the list) would be greatly appreciated.

Ok, what I designed the top navbar for originally was to list the main
sections of the site, and then the sidebar could take care of navigation
on the levels below that. It's similar to what we do with the tabs on
www.ximian.com, and lots of sites do this, it's almost become a standard
navigational device.

 
>> It's tempting to think about some changes in the look itself.

> Yes! :)

Of course, if I do something with this, that'll suck for Steve, since he
has templates and everything all worked out.

What I'm thinking about for the look is to move to a slightly cleaner
look, the scanlines in the top image plus the foot with the drop shadow
plus the beveled colored line and so on make things very busy, I think,
there's a lot of stuff going on. I'm thinking about something with
cleaner, single-pixel lines. But I'm not quite sure yet, I'll have to do
a little work on this.


>> (At this point, as many times before, I wish GNOME had a logo that was
>> more graphic design-oriented, and not so photorealistic and shaded. But
>> well.)

> I've been talking to tigert about this - not sure how much time he's going
> to have before 2.0 to work on it though.

Oh, that's interesting. The problem is that if we do this, people will
scream and shout because a) There wasn't a contest, b) A portion of the
people will always think any change is bad, and c) The marketing people
have time and effort invested in the current logo. So I don't know.


>> There's probably more stuff too, but I want to get some feedback on my
>> thoughts. As mentioned above, I'm willing and ready to put work into this,
>> to the point of implementing it all myself if everyone things these are
>> good ideas, but decide they don't want to do any work. :)

> I think the best spots to start banging on are IA and design issues,
> especially the correlation of the two. That would really help.

Ok, that's what I'll do, then. Plenty of stuff to do there, I think.

-- 
 Joakim Ziegler - Ximian Engineer - joakim ximian com - Radagast IRC
FIX sysop - Free Software Coder - Writer - FIDEL & Conglomerate hacker
          http://www.avmaria.com/ - http://www.ximian.com/