Re: Thoughts on 2-way RPC

Charles Goodwin wrote:

>Just thinking about it, obviously any RPC has to be client-initiated
>given the awkward locations a client might find itself in.
>  
>
Of course.

>Therefore, any form of 2-way RPC over XMLRPC will have to be
>polling-based.
>  
>
Not strictly true. Anytime you have the ability to keep a tcp session 
open (e.g. telnet or ssh), the server can initiate a transfer to the 
client without the client polling.

>So, well, we have an xwt.util namespace.  Let's add the 2-way RPC there
>and create a mini-standard of our own for interacting with it from the
>server.
>  
>
Check out this post from March of 2002 which talks about using http 
keep-alive to handle this reasonably efficiently. As far as I can tell, 
all of the mentioned issues are still quite relevant to xwt today.
http://lists.xwt.org/pipermail/dev/2002-March/000039.html

-Tupshin