Look, I get a laugh out of reading the announcements of new (or rebuilt old) operating systems. I am frankly pretty skeptical that world needs a great many more operating systems now that it has 1 poor, big commercial, closed-source one, 1 good free, small-or-large, open source one, and 1 good, small-or-large, half-open-source-half-closed-source one. From a market standpoint, I'm not sure we need more since it's really unclear where the gap is that is going to be filled by something else that can grow into something better than either of these; the alternative of replace-all-at-once cannot even be accomplished by one of the largest, richest companies in the world and is, in my view, too much work to be useful.


So, this new Phatom OS would have been just another laugh-at-somebody-else's-expense... except something caught my eye.


The clever bit, at least to me when I read it, was that by targeting only virtual machines you can actually produce a situation where the underlying operating system doesn't matter much. This is clever and appeals to my sense of the macroeconomics (see above) since it allows you to get something small out the door, get some customers on it for some special/custom operations, and grow into something bigger.


I have somewhat reached this agnostic point already. Nuxeo is a large system, but it runs on basically anything that has a decent JVM. There are some "environmental" issues with windows that make it a bit tougher to support than Mac or Linux (either 32 or 64), but from the standpoint of 99% of the development, the underlying OS doesn't matter at all. So, all-of-the-sudden we have myriad gaps in the market! If the OS runs the JVM well (or the CLR, if you care) you can switch to it. That is the only thing that really matters to me anyway, and with the JCK it seems feasible that this could be guaranteed to a reasonable level of compatibility.


Once this is achieved--and it's not trivial--you can start differentiating your OS with extra Java packages that exploit super-cool-feature-X. Who cares what it is, it is something that might entice someone to stay on (get stuck with) the OS. Similarly, if you support some weirdo hacks in your OS that allows some JVM operation (or, better yet, a whole method or class) to run 10X faster, then you might be able to differentiate on performance. (This is risky, though, because the linux weasels will steal it in 10 mins! Look at the Tomcat native support if you need any more proof!)


This idea should have occured to me before. It strikes me, having thought through this now, that it might be the sole reason anybody wants Solaris anymore.