Monday, May 5, 2008

Render unto ARC the things which are laggy and to Showcase the things that are way cool

Just read great posts on Rheta Shan's and Ciaran Laval's blogs related to new Second Life "features" showcase and avatar rendering cost (ARC.) The issues for me with both ARC and the rating system are relevance, accuracy and fairness. Rhetta brought up some interesting questions related to the relevance and accuracy of the ARC:
...the ARC colour rating is entirely independent of the power of your graphics hardware. How well your GPU would handle the workload never goes into the equation.
I think the idea of the ARC makes sense. I use the Mystitool HUD which has many features including automated scanning for nearby avatars. Some venues are set up to detect the HUD and request that it be put "to sleep." It doesn't bother me that a parcel owner limits lag-inducing items. So if there was an accurate ARC, it seems that it might make sense for parcel owners to be able to set some sort of upper level per visitor. I would personally like to know the real and relevant processing impact of objects I may want use. That said, bad data is much worse than no data at all.

On the subject of switching search rankings from traffic to some sort of editorialized "best sites" ranking, the issue is fairness. Even if the system would be equitable, I think the appearance of fairness is just as essential. Ciaran wrote:
The first thing that should be made transparent about this, the very first step to retaining business credibility, should be that no Linden Lab employees, their family members or their close personal friends are promoted via this process. The sort of thing that happens when companies run competitions, employees not allowed.
I've posted before on issues related to ratings and search rankings. So I'll just add here that I'm not opposed to some sort of editorial recommendation system, but I think it should be supplemented by traffic (paid users to minimize spoofing) and perhaps another shot at voting (one vote per paid user.) Another possibility (having just read "Little Brother," Cory Doctorow's great new book) would be some sort of Bayesian algorithm to weed out high probability bots.

Anyway, that's my two pence for the day.

1 comment:

Anonymous said...

Hi Botgirl, I started a comment that got really long winded and too long so I made it it's own post on my blog!