Re: The state of our web site and standards
- From: Jeff Waugh <jdub aphid net>
- To: gnome-hackers gnome org
- Cc: gnome-web-list gnome org
- Subject: Re: The state of our web site and standards
- Date: Thu, 12 Apr 2001 22:45:31 +1000
[ Cc'ed to gnome-web-list to continue discussion there. Webby dudes, you may
want to follow this discussion in the gnome-hackers email archives. The
first mail in the latest part of this thread can be found at:
http://lists.gnome.org/archives/gnome-hackers/2001-April/msg00057.html
... it introduces a few things we've discussed, and some we haven't. ]
<quote who="Daniel Veillard">
> On Thu, Apr 12, 2001 at 10:07:15PM +1000, Jeff Waugh wrote:
> > True, I did think about this whilst writing that email. I will suggest that
> > as pages are committed, we have a validator run over them automatically.
> > Kind of like tinderbox, but for web weenies. :)
>
> which is one way of doing it, late, i.e. once the person who
> did the job has commited.
> When you commit C file, any sensible person compiles and test it before
> commit, I don't see any strong enough argument why HTML authoring would
> be different from C authoring in this respect.
True enough. I'd imagine that the more active web hackers will run their own
local setup to to hack on, though. [ continued... ]
> The horror is putting an arbitrary change between the consumer and
> the producer, forcing the latter to generate something which is in an
> uncheckable format. Moreover the solution suggested requires rewriting
> of the content for every html page server, bad for the server and debugging
> IMHO.
Yes. For casual authors, you are quite correct, and this is something we
shall have to consider carefully, as we'd like a very wide audience of
possible contributors.
> > The web list has just picked up again, and it looks like it's time for the
> > dreaded implementation argument we've all been trying to avoid... So, if you
> > have any axes to grind (or preferably wisdom to share), get in quick.
>
> My suggestions are:
> - keep it simple and do it fast
> - provide templates for the various areas of the web site
> - serve static pages
> - provide tools to the authors to check their pages
> - forget about trying to gain 20% bandwith when serving images
> proper cache support and a single format is better (we used
> HTTP content negociation of images types at W3C it's hard,
> and not worth it unless you want to pass a political message).
>
> We still fon't have a link to the Foundation pages on the top Gnome pages.
> A lot of the content on the developper site will actually be generated
> though XSLT from docbook, the PhP trick will not work for it (we will
> use xsl:output="html" and the html struture from the stylesheet).
This is something that has been troubling me a bit, too. All of the
generated documentation (there is - and should be - *lots*) will fall
outside of the scriptable templatey stuff that we have been talking about on
gnome-web-list.
Perhaps a hybrid dynamic-static system will be required. One site I worked
on required a scheme like this, so we ran normal mod_php for most of the
site, and automatically spat out static pages for the documentation tree
using DocBook and a php-cgi script (just to keep language consistency, the
tools were not of my choice).
This meant we could use exactly the same php-based templating techniques,
but could also handle static content generation (by far the bulk of the page
views, so it had to be fast) and difficult-to-integrate content fromother
sources (in this case, as well as Gnome's, DocBook).
Might be some fun-ish design and hacking in that. :)
- Jeff
-- jdub aphid net --------------------------------- http://linux.conf.au/ --
ASCII stupid question, get a stupid ANSI.
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]