When you go into production
I love that I've gotten a bunch of non-IT-diehards/home-users/hobbyists/hackers/whatever to use the term "production." ;-)
Anyways, the long and the short of this all is that, yes, we know other things "work" - often for values of "work" that include "can fail, unexpectedly, catastrophically."
I've run FreeNAS on 1GB of RAM on a virtual machine and y'all can suck on that.
The whole point here is not that it CANNOT be done, as
@pirateghost points out, but that it SHOULD not be done. If you are here, on ZFS, on FreeNAS, we assume that it's because you value your data. We do not want to be responsible for guiding users to catastrofail solutions. If I give advice, I will give advice about things I'm confident in, usually things I've actually done.
We also strongly discourage users with anecdotal evidence of success from promoting solutions that are known to fail, without providing a warning to that effect.
So I watched very carefully for a year as we saw 4GB and 6GB configurations that failed mysteriously for people, resulting in some pool losses, instability, etc. Many others worked just fine, I'm sure. I was the author who changed the manual, upgrading the minimum RAM required from 6GB to 8GB. We still don't know exactly why less didn't always work correctly, but the evidence was that the problems vanished at 8GB.
My beef today and the bullshit I was calling is the hostile experiance you provide to anyone that does not shell out the $$ for server hardware for a movie server.
We understand the problems with ECC, or with Realtek ethernets, or any of the other things we like to harp on. We're not necessarily saying these things won't work, but there are known issues, for certain. You can absolutely make a FreeNAS box run on a random pile of crap (* note, no longer x86) but the results may be less than spectacular, which is why some of us have spent so much time and effort trying to document what DOES work well.
So please chill out.