begin quoting Andrew P. Lentvorski, Jr. as of Wed, Mar 23, 2005 at 03:59:55PM -0800: > On Mar 23, 2005, at 10:53 AM, Stewart Stremler wrote: > >>think it has SMP now although probably not as mature as Linux's SMP > >>since it has been around so much longer. > > > >By this argument, I should jump on the OpenSolaris bandwagon. Linux SMP > >support has never been all that great. > > <grins sheepishly> > > Uh, yes. In fact, I have been looking at moving quite a lot of systems > to OpenSolaris. Heh. Didn't mean to drag your intentions out in the open there. :)
> The SMP support is one of the big features. However, the big feature I > am looking forward to is: > *quality assurance* Very important for business. Far less important for hobbyists. Home users fall somewhere in between. > The simple fact that someone, somewhere has a set of programs that > *must work* with the new version with no changes. The fact that > someone tested this. The fact that I can get the *exact* configuration > as I had 4 years ago *and it still works*. Runtime compatible, our just source-compatible? > I have stuff that was written and compiled for Solaris 2.6 Sparc. It > still runs on Solaris 2.9 Sparc (we're testing on Solaris 10 Sparc now, > but expect to see no failures there). That's almost what? 10 *years*. > Try finding a Linux binary from 10 years ago and see if it runs on > *any* Linux today. If Sun can start this kind of support in the x86 > line, I will be hard pressed to recommend any of the Linux or *BSD OS's > as a corporate solution. I tend to hold the position that the benefit of cheap x86 hardware is offset by this lack in the x86 operating systems, especially for the corporate world. The amortized cost of the hardware isn't significant compared to the costs of "keeping up". Then again, there are programs that are written for a specific version of Solaris... I ran across a few in the 2.5/2.51 days, and most of 'em I fooled by replacing uname with a wrapper that checked a magic environment variable. I dislike code that checks the system version and aborts if the string isn't exactly what it wants. It doesn't matter how good the OS is if the developer decides to lock in to a particular versions. (But that's another example of 'hard-coding'...) > FreeBSD used to be the only thing that even *tried*. For a while, you > could run early 2.X stuff in 4.X series. I think most of that broke as > of the 5.X upgrade. I really hope they get back to that attitude of > providing backward compatible shims. Shims, or thinking ahead in the design? -Stewart "A fan of IFF (Interchange File Format)" Stremler -- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
