Pimp my URI

February 3, 2009 12:00 PM

Flat- or tree- URL scheme for a site redesign?

I am in the process of migrating a site (from Typo3 to Django, yay), and need to make a decision about URL schemes.

The site previously had URLs of the kind: "/category/name_of_entry". However, the categories have moved around, with old ones being deleted, new ones created, and entries moved as necessary. This will ikely continue to be the case. Tim Berners Lee says that you should never, ever change the URL of a page, and I see his point.

So I'm planning to use a flat URL scheme, sans categories, something like "entries/name_of_entry". This implies changing the URLs at least this one time. I also seem to remember that the googleplex likes sites with a treelike structure, would throwing everything in a single non-category affect their opinion of us?

I might use date-based URLs, but I don't have this information for the few hundred existing entries. We have 'editions', so I could also set up the URLs around them, but it still seems kind of messy, as some articles don't belong to any edition.

I'm thinking of setting up a redirect so the old URLs still resolve, so we don't lose all inbound links.

What are your thoughts, best practices and advice on this?

posted by signal to computers & internet (7 comments total)

Best practice is almost always to make URLs meaningful. Date-based URLs are not ideal unless the corresponding page is specifically about that date (e.g. a diary or daily blog).

Do try to preserve your existing URLs. If that means mapping the legacy URLs to your new scheme somehow then try to do that.

The old scheme /category/name_of_entry sounds perfectly sensible.
posted by le morte de bea arthur at 12:15 PM on February 3, 2009

I personally like to organize things in directories using a tree-style hierarchy. If you're the only one running the site (and thus you won't have anyone fighting over structural organization with you), I'd do that.

For sites that are collaborative, like Wikipedia, I can see going with a flat structure, because people are capable of wasting near-infinite time arguing over hierarchical categorization.
posted by Robson at 12:15 PM on February 3, 2009

The old scheme /category/name_of_entry sounds perfectly sensible.

It would be perfectly sensible if the categories where stable. They're not, they change over time, so the URLs would change, which is what I'm trying to avoid.
posted by signal at 12:29 PM on February 3, 2009

I like the way valleywag does it, with http://host/unique id/title. The post lookup is done by the unique id, so if the URL gets truncated or changed, the page is still served, but the title is there for people to look at. I have no idea about the SEO implications of this though.

see: http://valleywag.gawker.com/5131545/metafilter-users-debate-url-practices
posted by doteatop at 12:31 PM on February 3, 2009

Sir Tim Berners-Lee has said quite a few things. And I don't think they're gospel. Taking each issuance from the W3C as something to which someone must adhere is the Biblical literalism of web architecture. You could end up with large, crufty nests of redirects in an Apache configuration file two decades from now. Site URLs could never change, you could only leave 301s up, forever.

Keep your inbound links, but only if you have hits on them, and redirect with an HTTP Status of 301. Review on a regular basis — once the hits drop off, remove them. After two years, drop them.

I'm not sure about the tree-like structure. If linkfarms store things in a flatter structure, then go for a tree.
posted by adipocere at 12:38 PM on February 3, 2009 [1 favorite]

The only person in the world who cares about the organizing rationale of your website is you. Everyone else just wants to find interesting content. So the folks who say never change a URL and make your URLs maximally meaningful are giving you great advice. I'd skip both /category/ and /entries/ -- they're just more keystrokes that can be typed incorrectly, and they neither assists visitors to find something worth their attention.

Google rewards websites that appear to be genuine collections of information put together by real humans with a texture of incoming and outgoing links that look like real relationships. The most crassly scripted webscraping pantomime websites may at some time have looked flatter and less heirarchical, but I don't think that's necessarily the case anymore. Your genuine website will get flagged as genuine regardless of its architectural shape.
posted by gum at 12:46 PM on February 3, 2009

Hierarchical structures, despite their beautiful and soothing order, seem to be going the way of the dodo, in favour of tags and more dynamic representations.

Your site, with categories being created, changed and deleted constantly would seem to be a great fit for a flat structure, or at most, a very limited handful of meta-categories. These would have to be absolutely immovable and unchangeable.

Thus, your URIs would keep the "/category/name_of_entry" scheme, but you would have no more than 5 to 10 categories. All other categories would become flexible and perishable subcategories and would be handled as tags.
posted by Cobalt at 5:12 PM on February 3, 2009