polling only on socket, or can there be a trigger?

[Mark Waddingham]

> My sequence is that the client connects to the server, and then sends 
> it
> periodic messages.  On the initial connection, the server creates a
> database connection and leaves it open (opening a database takes time
> measured in hundreds of milliseconds).   The approach of opening and
> closing a database on each transaction simply isn't a realistic option 
> for
> me.

All the socket commands can be used in a non-blocking, message-based 
way.

When you 'accept connections' you create a listening socket whose 
purpose is to wait for connections and then supply you with a 'normal' 
socket to talk to the client which connected. The listening socket will 
continue to persist after this (until you close it) continuing to wait 
for new connections from clients.

So an outline of an example (which assumes on receiving data from the 
client you need to send it something):
   on startServer
     accept connections ... with message "newClient"
   end startServer

   on newClient pClientSocket
     read from pClientSocket ... with message "clientDataReceived"
   end newClient

   on clientDataReceived pClientSocket, pData
     ... process pData ...
     write ... to socket pClientSocket with pData with message 
"clientDataSent"
   end clientDataReceived

   on clientDataSent pClientSocket
     read from pClientSocket ... with message "clientDataReceived"
   end clientDataSent

The idea here is that each time you get a 'newClient' message you have a 
newly named socket 'pClientSocket' which is the other end of the 
connection the client initiated. You can have as many of these 
client-connected sockets as you like, and independently read / write to 
them with messages sent when the action you request completes.

If you use the blocking form of the socket commands then (as they don't 
dispatch messages whilst blocking) you can basically only talk to one 
client at a time. The blocking form may well be fine for a client, if it 
is connecting to a server, interacting with some sort of linear protocol 
and then closing the connection; but for a server that wishes to handle 
multiple clients simultaneously it is a much better idea to use the 
messaging form.

Hope this helps,

Mark.

-- 
Mark Waddingham ~ mark at livecode.com ~ http://www.livecode.com/
LiveCode: Everyone can create apps