Hi,

I am keen on reading an answer to this answers.ros.org question:
http://answers.ros.org/question/2803/when-should-i-send-geometric-information-over-a
"When should I send geometric information over a topic as opposed to TF?"

I would think that TF is the best place to publish frames, meaning 
coordinate
systems to be used as a point of reference for other geometric data.

So this would be ideal for robot joints/links, sensor coordinate systems,
and perceived joints/links of other agents (humans or robots).

If possible, all those should rather be published in TF rather than 
anything
else, I would believe.

However I do not yet understand whether there is a downside to using TF, or
some kind of degradation when plenty of frames are being published.

The only thing I can think of is that rosbag and rxbag cannot treat 
individual
transforms, they can only record / replay the whole tf tree.

Also I guess that data with a lot of noise is not suitable as a 
reference frame
for anything, so there is no point in publishing that in TF.

Other thoughts?

cheers,
   Thibault