Netcraft -- er, Jim Zemlin, confirms it: Solaris is dying. Customers are leaving it and legacy Unix behind for Linux, in his purview. Open sourcing the platform was too little, too late. Well, maybe not sundown, but it's getting mighty dark out.
These are actually not new sentiments; I picked them up from Jim when I talked to him back at OSCON -- a place where, ironically enough, I had also talked to folks from Sun. They were and are smart guys, deeply proud of the work they're doing, but I hope they all understand they are never going to steal any of Linux's thunder. (The refrain I've heard from many different quarters about this issue has been expressed in almost the same exact words by all concerned: "If only they had done this [open sourced Solaris] three/five/ten years earlier...")
If there is one central reason why Linux has done such a great job of eating Solaris's lunch, dinner, dessert, and midnight snacks, it can be summed up in a single word: agility. Drop Linux most anywhere and it lands on its feet, thanks to all of the different implementations that allow the kernel to be used in everything from embedded devices to clusters. Solaris's biggest selling / usage points are things that still mainly appeal to programmers (Dtrace) and people with high-end storage (ZFS) or processing requirements -- and in licensing that's not compatible with Linux* on top of that.
That also means Solaris is not going to vanish overnight, for many of the same reasons legacy Unix hasn't completely disappeared yet. There are still plenty of scenarios where Solaris is useful. The question is, a) how long will it be before all of those scenarios are eclipsed by some implementation of Linux, and b) will that happen at a total cost highly competitive with Solaris?
Given how good a job Linux has done getting a toehold in most every other place (except maybe the consumer desktop, and even that's changing, thanks to the netbook market), I'd say a) soon, and b) heck, yes.
If Solaris has a future, it isn't going to be in doing what it's always done, and that's bigger than just opening up the code and expecting a community to come to it automatically. And if Sun's people are wise, they won't wait around for hard(er) proof of that.
[*Footnote: From what I've been told, it is possible to run ZFS in userspace on Linux and thus work around issues of licensing -- but it's still pretty disadvantageous (shilling for disingenuous) for ZFS to be licensed in a non-GPL fashion.]