good idea, but I doubt that would happen. The UD.exe is tagged as copyrighted/trademarked to United Devices, Inc,

IBM doesn't own the UD.exe that you would need the source to port to linux/mac. It's not their intellectual property to give away.

Tossing in my own 2 cents, I suspect that WCG will spend more time on the agent design than on the actual coding. Once they decide just what the design should be, it will probably not be very difficult to port it from one OS to another. One thing we are running into with the current agent is that it repeatedly accesses the server trying to continually download new Work Units if something is wrong, rather than stopping and displaying useful error messages. People keep wondering if results are getting back to the server because the Agent does not have any log or timestamp showing when the latest result was successfully returned. The current Agent is not well designed for mass deployment.

Yes, I am happy that we have so few problems. However, a new Agent for Linux allows us an opportunity to correct the problems revealed in the current Windows Agent. And once the new Linux Agent is live-tested and debugged, we can begin thinking about eventually porting it back to Windows.

My main point is that we should not rush out the next Agent but instead spend a lot of time on the paper design, before we start coding it. There is always a tension between getting something out and getting it right. We have correctly started by building the World Community Grid on an existing plan. But with this experience, we can see where improvements would be helpful for the next Agent.

Improved error-detection would reduce the burden on the servers to continually download Work Units to faulty device installations. And improved logs would reassure new members and aid in error identification and correction. Finally, the current installation program is heavily oriented toward consumer single-computer installations, which is sensible. But for long term growth, we need to make some provisions for mass installs, such as in an entire school.

At the same time, we need to be wary of putting too much into the Agent and carefully consider what should be left for higher level supervisory utilities. I have already stated my opinion that 'process throttles' are really an OS function for controlling a CPU and its operations. If the OS does not have any native controls, then probably an add-on utility would be the way to go.

Hi Frogger01

I'm not suggesting that Linux cannot calculate information any less accurately than Windows. The point that I am trying to make is that the results of this project have life or death consequences at the end of it.

As such changes to the code need to be performed incrementally in a controlled environment at IBM by Rick Alther and his team.

Having the program as open source would allow anyone to change the client and therefore the accuracy of any WU results would be called into question on each unit returned.

By doing so this could invalidate all 2000 years+ of our contribution to the effort so far.

Regards

Dave

[edit]
Change of my post so that it does not misrepresent "reality" (as defined by Alther)
[end edit]