Are SES URLs evil?
I was reading a recent article in .NET magazine about optimising Flash sites for Google. When I got to the box on Friendly URLs, I thought it ironic that they were perpetuating another web design myth. After all, their recent June issue had been dedicated to myth busting, and SES URLs are the biggest myth out there?
Checking back however, I realised that even .NET magazine are still taken in …
Thanks,
Rob
a) IIS supports SES URLs, and you can, for the most part, uses almost the exact same configuration at the server. I develop on Apache at home and have no problem moving to IIS in production. There are multiple IIS plugins that support SES URLs.
b) Relative URLs - to me, this isn't a huge deal. Where I type ../foo.cfm or /foo isn't any harder or easier. I'd maybe argue that it's even better. If I'm writing a link from some resource, and that resource moves, then I'd have to update the link. (Of course, if foo moved I'm in the same bot.)
"You need to use a rewrite module such as Apache's mod_rewrite. As well as the extra time and knowledge setting this up requires, it makes your application less portable if you want it to work on IIS as well."
ISAPI_rewrite for IIS uses apache's .htaccess files and syntax. Quite portable. Besides that, when moving a site, should copying rewrite rules really be such a large chore?
"Relative URLs don't work - you need to link everything from the web root. There's a lot more work involved if you wanted to change your directory structure."
If you're rewriting URLS, you don't really have a directory structure. You just have a few templates that are used.
"You need to take extra care not to duplicate content."
Lazy or bad programming is not a case against SES URLs.
"You also can't put session or other client parameters in the URL, or it will confuse the crawlers."
You certainly can put url variables in SES urls. Following your example above:
www.mysite.com/product/productID/123/?CFID=1234567...
Finally...
"Now many people mistakenly believe that search engines such as Google can't index pages with these URLs. Often they will re-write this URL along the lines of http://www.mysite.com/product/productID/123.";
The point of SES URLs is not to remove the ? and = characters from the url. It's to create human-pretty and machine-intelligent urls. Such as:
http://www.mysite.com/products/tdk-80mb-cdr/ (equivalent of http://www.mysite.com/product.cfm?proID=123)
into http://www.mysite.com/product/productID/123 really doesn't achieve much.
A better idea, for SEO and for human readability, is to rewrite to something like http://www.mysite.com/product/Xbox-360-Premium-60G... (you may or may not want to include the product ID in the URL, but my point is that the text is the important part).
Another thing is that it doesn't really matter if you can hit a page by multiple different URLs if you specify a canonical URL on your pages: http://googlewebmastercentral.blogspot.com/2009/02...
The point for me is that your URL schema is an interface to your application, and needs designing just like any other interface. It should ideally be dictated by an information design approach, rather than a particular framework. It's not even for the sake of search engines, which have, as you say, got much more intelligent in the last few years, it's for the sake of your users. Of course, you have to weigh the cost in terms of dev effort against the benefits, but i believe the benefits are worth it.
When I'm typing into my browser URL bar, trying to remember what URL it was that I visited for 20s two days ago that I'm trying to refer a friend to, it's SO much easier if the auto-suggest is showing me URLs like blah.com/articles/why-x-is-the-new-y-and-z-is-dead, or gruntfuddlr.com/products/mang_wurdler_3_42 as opposed to dogbotherer.com/index.cfm?fa=articles.show&article_id=125222. It also helps increase your search ranking if the term someone is searching for is actually in the url of the article.
It's not even that much effort to put an apache rewrite together with a simple translation layer on top of a Fusebox or Mach-II app. On the last big Fusebox site I did, we did something roughly equivalent to Rails' routes.rb in a couple of days, and it made a huge difference to the usability of the site.
Give it a try and see!
Al
Was this myth created by the ColdFusion community?
The point here, as most of the other commenters already said, is that "pretty URLs" have very little to do with search engines and more to do with building better, cleaner web apps.
I realise that setting up SES URLs aren't a huge amount of work, and I already acknowledged that there may be a usability benefit.
My main argument though, was that the web already has a well defined way to pass URL parameters. Passing this information via folder structure is a hack.
Now Google can easily recognise normal URL parameters, but it can't work them out if passed via folder structure. At the moment, Google only uses this information to work out which parameters to ignore, but in the future it may use them positively too.
In future Google and others might be able to work out what URL parameters do on your site automatically, and use them to build the next generation of screen scrappers. For example, it may work out that your website is a booking engine or quoting tool, and automatically interogate it as if it were using a public API.
Certainly a whole folder dedicated to any subject would be worth more (have more information) than a single file. So why would google not give you credit as such? It's just another step higher. For example:
/buick/ is better than /buick.html is better than /cars.cfm?make=buick
I've been on a few SEO courses, and it seems that a lot of the techniques people recommend are based purely on guess work.
Google doesn't give away it's algorithms in case people exploit them, so SEO experts rely on anecdotal evidence. I haven't heard of any scientific SEO tests that have been done, and I'm not sure they are possible in a real-world situation.
Ultimately, we only have Google's own guidelines to go on, and they don't mention a folder structure as having a hierarchical impact on SEO or any other SEO impact.
If Google recommend NOT using SES URLs, then that's what I'm inclined to go for.
I've seen sites with lower pagerank come up higher in results because of their content. PR is important, but that's more of a popularity number, which has been made to be less important than it used to be.
Gareth,
I don't have any control tests. But using logic it's hard to argue that friendly url names with a visible hierarchical structure must help SEO. For instance if someone is looking for a "compact florescent bulb", which of these is better:
a) /phillips-par20.html
b) /bulbs/florescent-compact/phillips-par20/
It's not just slashes for the sake of breaking up words. It's categorization of information.
You're right in that it's guesswork. But as long as it works, we can also call it witchcraft :)