Thank you for raising so many interesting points. At the same time, I am realising that the discussion is now getting much larger in subject than just on System-on-Chip. If you don't mind, for the SoC Interest Group and hopefully a presentation at the next meeting, that will be made by one of your colleagues (Aleksei?), I would just rather limit it on how to use the SoC in the best way within the given context. - For a larger discussion, we can continue using this email thread, and I am adding my colleagues from the xTCA support at CERN in CC.
Now, on the larger subject of infrastructure, here are a few comments:
* ATCA is IMHO a very badly written standard, but it provides us with a common hardware platform for the phase-2 upgrades of ATLAS and CMS. The standard, in particular HPM.2, has a clear distinction of platform management, i.e. the ATCA/IPMI, and the payload, which is where the SoC comes in. Following those ideas, there is separation of control and data, even if those data in the end may contain another, higher level of control, purely from within the payload.
* There is one very important aspect, which I am not sure it has been given enough importance yet, and that is long-term maintainability. One of the strategies for providing that, is to push for commonality, of course, within reason. It seems to me that ATLAS and CMS are already largely engaged in using ATCA, and for the roadmap of some of the upgrade projects it would be too risky to change that now.
* Open-source for the core software certainly has advantages, but is not a panacea in itself. Commonality and long-term maintenance are certainly the key issues. As for the IPMC software, I fully agree that in order to ensure reliability, the deployments should be locked down as soon as possible in the development of a project. And that is very likely what we are seeing today with different projects being in different stages. I also think that both, open-source and commercial software, can be used to achieve this kind of locking.
I am happy to discuss this further if you wish. And maybe some of my colleagues from the xTCA support team also would also like to react.