I tend to agree with the lava article, although my reasons differ a bit. I seperated the GUI for one simple reason, Windows, to date, does not allow for a single workspace for multiple users. IT departments are getting more restrictive, and as such, require a user to log off a terminal when done. They are not allowing us to use 'common' user names as passwords. By having the test engine as a service that is launched as a system activitiy, the GUI 'plugs-in' to display the data regardless of which user is logged in. Thus avoiding issues with applications closing after a forced logout.
But at the core, the basic question of whether to use one, or more than one loop in a program really depends on what the programmer is doing. I have several of the Queued State Machines (QSMs) that are running as sub-vis, and as such are given to handling medium level tasks where decisions and order of execution play a part, but they never rely or include an event structure to operate.... Check that... I have one that reads data from a RS-232 port that is event driven, in that it waits (till timeout) for a set number of bytes at the serial port (event callback) before executing.
Of course many people smarter than I could hash these arguements out for a long time. My advice, be aware of different programming architectures and use what makes sense.
Message Edited by PJS on 11-05-2009 01:05 PM
Paul
) were lost in the LAVA crash of 2009.
