Suggestion for future client development

[John Naylor]

A lot of people on this forum (and no doubt other forums) have posted in frustration about the time it takes for new features to be added or bugs to be fixed in client development, and many have wondered why the project will not go open source... The answer will be obvious to anyone who has been on this forum for any length of time. However, I think that maybe people have looked at this from the wrong angle.

It is true that the encryption methods used by the client and the core need to be kept secret, which kind of necessitates closed source code. However it is also true that the Pande Group has limited time and funds for client development, which would suggest Open Source is the best method to go for rapid development. Would it not be possible to use both of these methods in client development? My suggestion is that the secure code in the client is seperated from the client into a seperate file (maybe a DLL) made available in compiled form only, with a clear, documented API on how to use it. The rest of the client, now stripped of the code needing closed source could then be made open source allowing for faster patching and quicker addition of new features, without the downsides experienced by BOINC in the need to repeat work to ensure it is correct.

I must make it clear that I am not a coder, so I have no idea if this would be feasible or not, and I guess given how long it has been since it was first said that v7 was in development that that must be well along the way to release. However, for future client development after v7, would this be a possible/viable way to go? Discuss

EDIT: I should add that this would not affect the FahCores, which would remain closed source by necessity.

[bruce]

I think the issues that you'd like to solve with open source will very likely be addressed as FAH transitions to the professional programming staff that Dr. Pande has mentioned in his blog.

[7im]

Even with Stanford's hard line against open sourcing certain portions of the client and fahcores, there has been mixed results on both sides when they tried to open up. They tried to open source the viewer, and didn't get any takers. Then others have attempted to improve performance by renaming fahcores, or running ATI GPU WUs on AMD hardware, or changing the makeup of the SMP fahcore, and each time corrupting hundreds if not thousands of work units.

IMO, the move to professional coders will slow things down even more, but then again, more stuff will be right the first time out, so no waiting for patches. Might eventually be quicker in the long run. Certainly cleaner and less buggy, and quicker to be fixed.

And anyone with the open source itch can read the FAQ about it, and channel the itch in to Gromacs.org, Amber, or one of the many 3rd party apps. http://folding.stanford.edu/English/FAQ-OpenSource

P.S. And special mention must be made of how quickly new features and fixes were made to the Systray gui when it rolled out with v6.x.