On Tue, Dec 11, 2012, at 11:32 PM, Marcus Sorensen wrote: > This is pretty important. Anyone should be able to roll their own, > rather > than relying on a single potentially out-of-date image. It seems like it > would be pretty simple and straightforward on the face of it, however > many > of the scripts have been written specifically for Debian. I'd honestly be > ok with having to stick to a particular distro if I at least had clear > instructions on how to make my own, I understand the need to program > against a single defined userspace.
I see a potential problem with this. Any scenario where users are customizing part of the stack means additional variables which means additional problems. If we target Debian, trying to create a system VM from CentOS/RHEL means different libraries, etc. - which means a number of potential problems cropping up where there were none before. I'm not saying users *shouldn't* be able to do this - just that I haven't noticed anyone raising the issue that we'll probably start seeing a fair number more bugs if replacing the system VM becomes a standard practice. There's a reason, for instance, that Linux vendors don't support custom kernels - and what's being proposed here is swapping out an entire OS. It's going to make things a bit more tricky when someone reports a bug and they're using a roll-your-own system VM and the people doing the testing are using a different one. Again - not saying we shouldn't do this, but I'd like to see that given a bit more consideration when we're discussing the issue. Best, jzb -- Joe Brockmeier j...@zonker.net Twitter: @jzb http://www.dissociatedpress.net/
