Folding Forum

In the lab I work at we're running BOINC for years. Being an engineering design lab, we have quite a few (100's) good specs CPUs and GPUs, so our contribution is significant I'd say.

Now recently a colleague suggested us trying F@H to support COVID, and I was the one with the strongest opinion against it - reason being very simple: it is a proprietary software.

It has been mentioned in the past, even quoted in Wikipedias F@H page that there's work ongoing to open the source or allow for open source plug-ins. Honestly the latter is out of interest for me, as the client being proprietary kills the deal. The reasons mentioned in WIkipedias to keep the source closed are, quote, security and scientific integrity. Well, not going into detailed discussion here on why these 2 reasons are simply wrong and having a proprietary software adds nothing to any of these.

Hence the question, to which I could not find a reasonable answer: what the progress of the work to open the source of the client, if there's any?

The project agrees that security through obscurity isn't sustainable. The work to open-source the client got sidetracked when Covid-19 happened and the single salaried developer had to be prioritized to keeping the work servers running. There are now two salaried developers on the project, but I don't know the progress on open-sourcing the client.

Both the folding cores (GROMACS and OpenMM), as well as the GUI etc. are already open-sourced, so it's just the client which is the sticking point now.

You might perhaps look into supporting Open Pandemic through BOINC in the interim.

As mentioned several parts of the folding client have been made open source. The developer was working on preparing the rest to be open source before COVID-19, he was tasked with doing one minor upgrade to the client to fix some bugs and add the COVID preference in 2020. The rest of the time since COVID has been working on the server code. They hired an additional developer about 6-7 months ago to work on the client code, version 8 may be out later this year.

Current open source elements are FAHControl, the Web Control, GROMACS, and OpenMM on - https://github.com/FoldingAtHome. Following links there also shows FAHViewer on - https://github.com/FoldingCommunity/fah-viewer. If I am reading things right, links to version 8 client code are also posted on the first site with '-bastet' appended to the names. Through one thing said over a year ago, the shipping client will be open source except for possibly a closed "blob" of code to handle the digitally signed files being downloaded and uploaded.

The only parts that are not open source are related to security and result integrity to prevent people from cheating (yes, it already happened).

The closed source parts are provided as libraries (functions) to be called by open source parts (client, viewer, ...).

BOINC relies on duplication of computation (quorum) to ensure result integrity, which is a waste of 50% the compute resources donated ... FAH makes use of all resources and only duplicates computation when an error is detected at one of various checks in the process. Those checks are provided by the closed source libraries. Don't expect this to change.

I have been led to believe that server-side validation would be enough to avoid cheating (deliberately trying to pass off bad data as good), but maybe the closed blob is there to avoid DDoS'ing?

I am not going to give advice on how to cheat F@H, but identity theft is a real threat if you are making executables separate from the F@H site. A easy cheat on SETI@home (preBOINC) was to compute 99% of a WU 99% and then give it to your entire team. (Another reason for F@H reluctance to issue a WU multiple times)

I am happy with not having any of this drama of F@H, even if we lose some open source purists. I prefer to be able to trust my fellow folders.