Re: wgo url policy

[Gergely Nagy <greg gnome hu>]

On Fri, 2006-07-14 at 00:24 +1000, Jeff Waugh wrote:
> <quote who="Dave Neary">
> 
> > > Minimal redundancy
> > > Avoid multiple URLs for a given page. Currently both http://gnome.org
> > > and http://www.gnome.org exist. I suggest to forward all gnome.org
> > > access to www.gnome.org. Are there any duplicate pages on wgo now?
> > 
> > These are not duplicate pages, they are aliases. gnome.org is an alias
> > for www.gnome.org.
> 
> That's not what Gergely was pointing out, though the proximity of the two
> points makes it easy to misunderstand his intent. First up, he's suggesting
> that gnome.org should simply redirect to www.gnome.org, such that all Google
> juice and so on points there, and we be very consistent in our use of the
> domain name (a very good point - and now fixed, by the way). Additionally,
> he was asking if there were duplicate pages (as opposed to domains) on the
> site.

Oops, yeah, 2 points in 1 mix :)
First, domain aliases are imho bad, because they introduce
inconsistency. Reasons:
- proxies and browser caches will not realize it's the same thing
- same with indexing engines (Google juice as Jeff points out)
- confusing for (novice?) users
- inconsistent links in documentation, about boxes, etc
- history, visited links in the browser will not work properly 

(btw i noticed gnomejournal.org has this aliasing too)

As a second issue I meant possible duplicate information within wgo.
Perhaps this better belongs to the "content mapping" task.

> > > - I suggest no extensions to file names, this way resources can act as
> > > containers
> > > for example:
> > > www.gnome.org/start and www.gnome.org/start/2.14
> > > instead of
> > > www.gnome.org/start.html and www.gnome.org/start/2.14 
> > 
> > I don't think this is really an issue. Direct linking to a html page is
> > common, having one directory per page (which is what would result) is
> > unwieldy, and (as you point out) .odf, .pdf, .jpg, etc - files have
> > extensions, those extensions convey some information. I don't think we
> > need a policy for this.
> 
> I do think this is an issue, and one that I have pushed as much as possible
> on wgo as it stands today. It is significantly clearer and more future proof
> to have a directory with an appropriately named index file in it than to be
> referencing specific files all the time. I 100% support the *continuation*
> of this policy. :-)

This important, since one of our goals is to increase consistency and
interoperability between individual gnome sites. Another goal is to be
CMS independent as much as possible, so we can possibly swap the
underlying infrastructure if a better comes along, without breaking any
links. Properly thought-after URL naming can abstract away
implementation details. I applaud the current wgo scheme :)

> > > I could not yet decide what I think is the best URL scheme for I18N.
> > 
> > I think the best is no URL scheme for languages and use the browser
> > preferences.
> 
> If you want to let people change the displayed language if the negotiated
> one is incorrect or not desireable, you have to select some kind of URL
> scheme. I don't think this needs to be fully defined ahead of time - the
> scope for this will be defined by the technology we choose, though we can
> certainly set some parameters. For instance: we can't use subdomains for
> language codes, because some subdomains are already in use by local GNOME
> groups; we probably shouldn't be screwing up our URLs by sticking language
> codes in there; perhaps the query string approach is the best (can you use a
> query string to override Apache content negotation already?).


I think the URL should be independent of the underlying technology, see
my point above. I just hope we can pull it off :) Good point on not
using subdomains, they can be reserved for local content (for local
groups and their local content, not the same general content translated
into a bunch of languages).

Like Jeff, I also tend to think that query string approach is the
cleanest. It implies it's the same resource with a different parameter,
that is "display language". Can be problematic though with some CMS
engines and usual URL rewriting rules (that for example would
translate /foo/bar/baz?lang=hu into /index.php?page=foo/bar/baz?lang=hu
which is obviously wrong: ? appears twice).

I guess we all agree, that a page in a specific language needs it's own
URL?



I realize I'm quite anal about getting the URL scheme right :) But I
think URLs are a very important and so far neglected aspect of proper
web presence (I don't mean the gnome web, but in general). For more info
see:
http://live.gnome.org/GnomeWeb/PlatformIssues
http://www.w3.org/Provider/Style/URI
http://www.w3.org/TR/webarch/
http://www.gbiv.com/protocols/uri/rfc/rfc3986.html
or even
http://www.ics.uci.edu/%7Efielding/pubs/dissertation/top.htm

cheers,
Greg