Use of absolute URLs and Swiss mirror

While the category pages, as well as SOME of the other pages are given as relative URLs in the HTML code of the dmoz website, (probably) most of the other pages are coded absolutely with the



This makes links unpleasant to use when visiting the Swiss mirror (which from here in France works, while the main site is awfully slow) : it is necessary to cut-and-paste them into the browser window, and add the "ch." by hand !



I could add that there is no apparent logic under the choice of absolute or relative URLs, e.g. on "help/helpmain.html", the URL for "Press Center" is coded absolutely, and immediately under the one for "Policies and Procedures" relatively, while both point to subcategory pages.



I suppose this should not be very difficult to mend.

Bonjour Frenchtourist,
Tu peux visiter le forum en français, il y a un fil sur ce sujet
si j'ai bien compris ta question <img src="/images/icons/wink.gif" alt="" />

You can visit the french forum, we have discussed it.

(Je ne continue pas en français, ce n'est pas la place ici !)

No, the thread you kindly pointed to me concerns the intricate problem of characters outside the ASCII range. The thing I was thinking is much plainer : this is the frequent use of hardcoded "http://dmoz.org" in the HTML sources, a consequence of which is that, even if the user was connected at ch.dmoz, when he clicks on the link, the page is fetched on dmoz.org (losing the mirror benefits).

Many pages on dmoz were created years ago when the directory was very small and there was only one person working on it. Some of the more unusual coding practises were put in place then and have not needed to be updated since.

The swiss mirror is the first such mirror, and it's there that many of these absolute links have been discovered. Staff have corrected dozens of such links over the past few months, and there should be very few left by now.

The exceptions are any links which perform actions. The "Add URL" and "Update URL" links still have to point to the original dmoz as that's where the database is. All of the editor links have to go to the original dmoz as well.

I think he may be talking about pages like http://ch.dmoz.org/Arts/Animation/desc.html

Oh. <img src="/images/icons/frown.gif" alt="" />

Those didn't ocurr to me. There's no way of sorting those at the moment. We'll have to look into ways in which they can be automagically fixed.

This isn't going to happen overnight though. Don't hold your breath.

If you mean things in category descriptions like "Do not submit blah-blah-blah kind of sites to this category; see Arts: Comics: Titles instead"... I've been guilty of that, and I was about to go and take off the "http://dmoz.org" from them, but there's a snag: URLs that start at the root (i.e. "/Arts/Comics/Titles/") will work with the mirror site, but may fail with sites that license ODP data, if they don't construct URLs in the same way. For instance, in the Google directory (I know Google doesn't currently show our category descriptions, but let's say it did) there's an extra "Top/" in every category URL.

So, to be as widely compatible as possible, you'd need to use relative URLs like "../../Titles/" depending on what category you're starting from. Of course there's no guarantee that ODP licensees will be consistent with us even to that degree, but it's a safer bet.

... and then there is the further problem that if you use relative links, then if the category is moved someplace else in the directory (either the one containing the link, or the one being pointed to) then the link becomes broken.

So, it looks like it isn't possible to fix this to work for everyone all of the time.

http://ch.dmoz.org/ no longer works. <img src="/images/icons/frown.gif" alt="" />

http://ch.dmoz.org/ (as well as all other URI on ch.dmoz.org) returns a 301 (Moved Permanently) and redirects to http://www.init7.net/ .

At this moment, http://ch.dmoz.org/ works...

Working fine.

Yup, worked fine last night, works fine this morning.

But I can confirm that it was down for some time yesterday - maybe a server reboot or something. Since all data is still there it is nothing to worry about anymore