[Verse-dev] Saving and Loading

[Samuel Siltanen]

On Wed, 14 Sep 2005, Eskil Steenberg wrote:

> Storing things like vertex count and and poly count and material count in
> tags is inadvisable. The problem is that this information is derived form
> data alreddy defined by the verse data set. If an application makes a
> change to the data set thees values become invalid, and since data in
> verse can change at any time this is not very robust. If you write a
> application that require thees values to exist or be correct, you will run
> in to all kinds of problems due to latency and your app will only work
> with a data set especialy prepared. It is much better to compute the info
> you need directly form the verse data.

I do not know if you are referring to my applications or if you are 
talking in general, but because it was I who mentioned this thing, I 
assume it is the former case. So I quote myself:

"These numbers are not strict boundaries, because one cannot expect that 
Verse would give a certain number of nodes, but they indicate when to start
processing the data further. This is not done until sufficiently many 
nodes are loaded."

Thus I do not assume that the numbers are correct. The application just 
waits for a certain time if the current number of nodes is below the 
target number of nodes. Otherwise it continues to process the data 
(subscribing to certain nodes etc.), but still accepts nodes that arrive
later. In some cases the application also keeps track of the traffic 
(callbacks called) and sets a timer which is reset whenever new data 
arrives. When there is no more traffic and the timer expires, it is 
assumed that the model is loaded. This approach can be compared to the one 
used by the VML-saver. It uses only the timer, so that the user have to 
approximate the network capasity and the amount of data on the server and 
set the timer value on command line appropriatelly. I just wanted to 
automate this process giving the user less to worry about.

Samuel Siltanen
saasilta at cc.hut.fi