
Linux only has a small percentage of the computing market, however Microsoft already considers it a major competition as the open source OS steals the hearts of many users. Following the hard numbers though, Microsoft also increases its market share on both server and desktop space with time. The only logical explanation is that Linux steals quite a market share from the traditional UNIX providers (SCO, Sun, SGI, HP, IBM). But only Sun seems to truly be in a real Linux trouble, as it is the one with a resistance to Linux integration to its full product range.
IMHO, if Sun can hang on long enough to deliver Solaris 10, I think it will be something like this:

All/most user-space programs, libraries, etc -- the user "environment" if you will -- will be common (and LSB- compliant - this is already a stated goal for Solaris) between their Linux and Solaris products.
The kernel will be the differentiator: you need super-SMP, hot-swap, etc? Plug in the Solaris kernel on SPARC hardware. You want lower cost, 1-4 way (server or desktop)? Plug in the Linux kernel on x86 or SPARC. But most everything else will be the same between the 2 versions.
Then they will have a "complete" offering that becomes mostly hardware independent (will anyone really try to do a 100-way x86 server?) and yields a common "look and feel" to the users and the developers (one common API for 98% of everything; specializations -- as they will always exist for any vendor -- for the bleeding-edge massive 100s of CPU systems).
And they'll have the complete range of hardware -- from 1-way x86 servers (like the new V60) and desktops ("Mad Hatter") to SunFire F15K and beyond "big iron" boxes. With N1 (as it develops) to help manage and integrate it.
If they can just hang on...