keitai-l@appelsiini.net writes:
>I didn't get it. The point is to squeeze the data *before*
>it gets to the client.
I took the point to be that one needs LESS data if one simply paints the
fovea rather than the whole retina as one need only paint the portion of
the scene (the "screen") that would hit the fovea if the eye were looking
at that portion. Hence one would need to send only the data that
represented that portion of the screen. So - LESS DATA. A genuinely
virtual screen. Other people (whose fovea was not being tracked by the
laser) would see a mess.
So this is a "single eye solution" (with a mod. for two eyes attached to
the same brain - I guess eyes generally go in the same direction).
That is where keitai come in as they have limited CPU, lower bandwidth and
are generally single person.
It is a third approach to be taken alongside wider bandwidth and better
compression (and faster decompression which is the relevant element here).
Latency of feedback of eye movement is the main problem I suspect as this
would have to be fed back to the server so that a different data stream
was sent (that represented a different part of the screen). I wonder what
kind of latency would be acceptable... Human eyes do not generally move
THAT fast I guess...
One should take everything one reads with a pinch of salt - and this
article did not seem particularly foolish.
(PS - the above might be ballcocks as I read the article rather quickly)
Nick
Received on Mon Mar 11 06:23:15 2002