Re: RFB Caching technique

[Vamsi K. <vkr...@no...>]

Hi Bob,

> 
> Currently the VNC RFB protocol does not depend on any caching at all

> except for perhaps the displayed image in the client.  The simplicity

> is quite useful in order to implement simple clients.  Hardly any 
> memory is required on the client side. 

I agree that VNC is a thin client protocol. But, by making features
like
caching negotiable parts in the protocol, we will have the flexibility
of using
these features and good performance on high end *capable* devices. 

> The non-caching approach makes 
> it very easy for one server to serve multiple clients at once.  These

> clients may come and go a different times.  How will the multiple 
> client scenario be handled with the caching approach?
> 

Caching will be designed to work on a per client basis. VNC already
maintains 
per client state for features like VIEW ONLY , encoding etc.
We will implement caching in the same way. The space requirement will
be ~24 bytes
per cache entry and is not a real overhead. 

> Has testing (or simulation) been done in order to evaluate the 
> real-world benefits of the caching?

We have a very rough simulation done and it is giving good performance
in real-world situations. We are facing some implementation issues in
combination with tight encoding (we have to understand tight
implementation 
and it really takes time :-) ). So our observations have been with
other
encodings and they are all really promising. We can give more accurate
readings in near future.

regards
Vamsi