Re: [syndication] More Ratings stuff

[Julian Bond <julian@netmarketseurope.com>]

In article <002001c09091$a121abd0$070d0dc0@monster>, Jeff Barr
<jeff@vertexdev.com> writes
>5. I don't know how to distribute the accumulator site. Someone has to
>   receive records, tabulate them, and make the results available.

If all you're after is an average rating, the only real problem is
normalizing the scale so that two aggregators can talk the same
language. If I provide ID, AverageRating, CountOfRatings, MyScale. Then
you can combine your set of data to produce a composite average in
whatever scale you like. 

But this is open to abuse because it doesn't say anything about where
the ratings came from and if it got circular, three people actually
rating something could appear to be 3000. 

>We also need to add some time-sensitivity to the model. I would like to
>know about news that is popular in a given (or just the current) day.

I think this is a nice to have, but I actually don't think ratings need
to degrade over time. I think this is part of the filtering /
categorization.  

>I'm not too excited about ratings on rankers. While its formally
>very cool, I think it is a lot of overhead for not enough gain. 

<AOL>Me Too!</AOL>

-- 
Julian Bond mail:julian@netmarketseurope.com
workurl:http://www.netmarketseurope.com
weblog:http://roguemoon.manilasites.com
ICQ:33679668 Tel:+44 (0)20 7420 4363
tag: So many words, so little time