The water used for chip manufacture is a very ultrapure water created through an involved process using mixed media beds, filters, and reverse osmosis membranes. The fertilizer would have never made it to the chip but would have likely fouled the ultrapure water production equipment as it needs repetitively clean feed water. The molecules in the water actually etch the surface of the silicone if they are not removed. - according to an ultrapure water production class I attended.

Ooooh, lot of different things going on here, I just typed a page of explanation, wondered if it was to esoteric, and then Omniweb crashed and I lost it all. That guidance is for medical device development. Very specific. Very complicated. Also very complicated is how to read a guidance in terms of current enforcement. They are written by lawyers and enforced in a particular manner.

I was limiting my observation to the much simpler case of an IT department suddenly deciding to mandate a switch away from Windows, or leaning that way, or phasing things in... whatever.

When I buy laboratory instruments, they come with a validated piece of software that's validated line by line for that instrument. I have never seen validated piece of lab software validated for anything but windows. It may exist but is very uncommon) The validation process is expensive. If I wanted to sub out linux boxes for my windows machines, I could do it if I validated the software for each instrument. First that would take a few years and cost me about $10 million on about $3 million worth of instruments. Second I would need tech specs for the electronic in the instruments that the vendor is not going to provide. Validation is only possible because vendors spread the cost over all their sales and they don't redo it often. I can't afford to change and I have much greater resources at my disposal than most university professors.

For instance, I have one instrument that was validated by the vendor and only runs on a 10 year old copy of window. The vendor will not upgrade the software because they don't want to revalidate the code. It's expensive. That's why a single user license for excel (non validated) costs $90 and a cheap single user copy of the cheapest validated lab software program costs $5000 even though they may do all the same calculations.

Another aspect of validation that just occurred to me is that once complete it can't be changed, so it may be inherently illegal to distribute the source code for a medical device or lab instrument, because the government would never be able to tell if you were running the validation version or a modified and recompiled version. That is just a thought... I can't think of a case where it has come up.

I largely brought this up because it is a frustrating form of mandated government lock-in, and because it is a reality that the OP will need to consider.

I did not say all research. I said government regulated research. Not all research is government regulated. Although as companies are farming out grants for studies to colleges and universities more and more university research will be regulated. He just needs to assess what goes on at his school. In my case if our IT department decided to switch away from windows, I would have to object and have him make an exception for my labs, because the validation of my software is regularly reviewed for the government submissions I make and there is no non-windows validated software to control my instruments. One for instance... The code of federal regulations sections on Good Manufacturing Practices for Drugs (21 CFR) and the section on Good Laboratory practices for non clinical studies require all data controlled by software to be life cycle validated software. This means you cannot do your calculation in RLab or some other open source package. It means your data cannot be archived in open source database. It means that all of your analytical equipment must be controlled by validated software. The entire concept of life cycle validated software precludes the notion of open source as the expense is staggering. That is why an instrument can be controlled in a grad school project with open source software but as soon as it is a regulated study for government regulatory approval you will have to replace it with validated packages in the high five figures.

You are going to need to assess the things that simply cannot be switched to open source and find out how they will be implemented in the new "mostly" open source strategy. For instance, almost all scientific instruments come from vendors that only have a version for windows. Since most government regulations for research, particularly for clinical, medical, and drug research, require a life-cycle validated software, there will be no open source software for these groups. Since your school is very small you probably don't have much technical research equipment.