Comments:

2. 

   • Sami on Thu 06 Dec 2007
   • 01:04:00 AM UTC

   Targets

   I agree, and we've actually thought about this before. When we first started working on WOT, we had a long discussion on the naming scheme we should adopt for our "targets". I mean, the algorithms we designed for computing the reputations don't really care what objects our users rate, it doesn't matter if they're websites or specific pages or something else. As long as they have unique names, we can handle it.

   At first, since we wanted to rate things on the web, it was obvious that we should use URIs. Not only do they uniquely identify all web pages, but they also have a neat property of being hierarchical. Our algorithms can traverse the name hierarchy and compute reputations for objects based on ratings elsewhere in the hierarchy. We thought it was pretty cool. Unfortunately, the more we thought about it, the more apparent it became that using the full page URL might not be ideal for our purposes after all:

   • Privacy. If the system computed ratings for every page, the add-on would have to send us the entire URL for all the pages you visit, just to fetch the ratings. I don't know about you, but I personally wouldn't want to install that kind of a tracking device on my browser. However, if we ignore the path and use only the hostname, the add-on only needs to ask for ratings once for each site. It's the difference between telling someone you visited YouTube, and telling them which videos you watched. As far as I'm concerned, the less information the add-on sends to our servers, the better.
   • Resources. If the add-on were to contact our servers each time a user moves to another page, I'd say running the service would require at least an order of magnitude more server capacity. By using hostnames, we can configure the add-on to cache ratings more aggressively, resulting in less server load and a cheaper bandwidth bill. This might not be an issue for Google et al., but at the time, it was definitely an issue for us.
   • Longevity. Anyone who has followed a bad link to a "page not found" error knows that URLs change quite often. We thought it might seem like a waste of time to our users to rate a page that doesn't exist tomorrow. Or even if the URL remains unchanged, the contents of the page might not. Sure, this applies to hostnames as well, but to a much lesser extent. Website owners tend to commit to their domains much more than to their site structure.

   Of course, these are only some of the things we considered. As it often turns out in life, no matter what we chose, some compromises had to be made. After some debate, we decided to ignore the rest of the URL for now and limit our namespace to hostnames. Sure, the granularity of our ratings suffer, because we don't rate every page separately, but I believe this was the best choice for our purposes at this time. This is not to say that we couldn't expand ratings to specific pages in future. There are solutions to pretty much every problem here, it's only a question whether or when implementing them becomes feasible.

   Thanks for posting though, we'll keep your request in mind.

   •  Login or Register to post comments
• 

  • 309east on Thu 06 Dec 2007
  • 01:47:05 AM UTC

  Compromise

  Thanks for the reply, here's a small rebuttal.

  Privacy -
  Maybe you can snag something from the social networking sites and toss in a submission / voting system with optional participation. If you see a specific article that warrants a ding to the publisher's reputation, you can submit the url. Allow other users to comment and either confirm or deny the reputation modification.

  Resources -
  With the voluntary submission that may or may not impact your costs.

  Longevity -
  If the article has a finite life at that site it's fine. Just have this submission and voting influence the overall site rating. Maybe you guys could go into more (optional) detail in your ratings, so you could see how many pages on a site had been flagged for some reason or another in the past.

  •  Login or Register to post comments
3. 

   • talkingtofu2 on Fri 07 Dec 2007
   • 10:19:54 PM UTC

   Subdomains???

   Taylor Satula
   I Aggree Like Myspace,Facebook Or Freewebs So By Page It Can Be Found And Rooted Out By Its Self

   Or Better You Could Rate It By
   example.blogspot.com
   or www.example.com/example/

   •  Login or Register to post comments
• 

  • Sami on Fri 07 Dec 2007
  • 11:41:33 PM UTC

  Subdomains

  We already handle subdomains. And yes, if we decide to expand the name hierarchy to the rest of the URL in future, then yes, we would handle them quite similarly.

  •  Login or Register to post comments