Re: Re: Future for Mathematica on Solaris
- To: mathgroup at smc.vnet.net
- Subject: [mg99938] Re: [mg99883] Re: Future for Mathematica on Solaris
- From: John Fultz <jfultz at wolfram.com>
- Date: Tue, 19 May 2009 06:49:32 -0400 (EDT)
- Reply-to: jfultz at wolfram.com
On Mon, 18 May 2009 02:31:43 -0400 (EDT), Dr. David Kirkby wrote: > Peter Orion wrote: > >> I was interested in purchasing a copy of Mathematica for Solaris, but >> have seen various things on the net suggesting the Solaris (and to a >> lesser extent linux) versions are more buggy than the Windows version. >> >> Looking on the Wolfram web site where one can request a trial >> >> http://www.wolfram.com/products/mathematica/experience/request.cgi >> >> I find one can't request a trail on Solaris from the web site. >> >> So I contacted Wolfram Research directly by email and asked about a >> 14-day save-disabled trail. I received a reply in a couple of days, >> telling me no trial version of Mathematica exists for Solaris. >> >> Given the Solaris version is more expensive than the windows one, >> there are various reports of bugs on Solaris, and one can't even get a >> trail version, it does beg the question whether Wolfram Research are >> serious about the future of Mathematica on Solaris. >> >> Peter >> > > Was this for SPARC or x86? > > When I was a lot younger, Sun were pretty much the defacto standard in > scientific computing. But now the SPARC processors are simply not up to > the speed of the Intel/AMD CPUs. Sun SPARC systems have their strengths, > but I don't think it is in scientific computing like it used to be. > > > So it would not surprise me if there was not a long term commitment to > Mathematica on SPARC and perhaps Wolfram decision not to make a trial > available is a result of that. > > I see Solaris x86 as having a bright future. It's much more stable than > Linux (stable in the sense that not there are not n-million > distributions all a bit different.) It is also very fast and growing in > user base now it is open-source. > Given > > 1) Mathematica on Solaris x86 does not support Intel processors - only AMD > 2) Intel processors are more common than AMD ones > 3) A fix to Mathematica to allow it to run on Intel processors is quite > easy (just replace a couple of libraries from those supplied by WRI to > those supplied by Sun) > > I am somewhat surprised Wolfram have not made more effort for Solaris > x86. So perhaps Solaris support will be dropped, like HP-UX support was > recently dropped. > > Having seem various talks from WRI staff, its clear that almost all > their staff's usage is on Windows. Unless more effort is made to get > their staff to regularly use other platforms, bugs are far more likely > to be found internally on Windows than on those other platforms. Hence > users of the rarer platforms are more likely to encounter bugs. (That's > my theory anyway - it might be completely wrong). So far as we know, 7.0.1 works out of the box on a Solaris Intel system. We routinely test Intel systems internally (especially once we realized that these problems keep cropping up on Solaris...FWIW, we *never* have had these kinds of problems on other OS's which run both AMD and Intel). If you're seeing a problem with 7.0.1, you should definitely report it to Technical Support. I've seen quite a number of Solaris issues that are specific to some combination of platform and OS patch-level, and it's very possible that you're running a configuration we're not testing. My observation, actually, is that Wolfram R&D have put in far more development time into supporting the various Solaris platforms in the last few years then you might expect given the proportion of sales that the platforms generate. I would absolutely not call the current state of the software neglected. It is certainly true, of course, that every platform Wolfram supports costs money, and quite a lot of it at that (far more than you would expect if you haven't worked for in software development before). Business realities can and do change, and when it doesn't look like a winner to continue to support a platform, it will get dropped. It's also true that the extras cost money, and that includes trial versions. Without projecting anything about the future (because I've got no crystal ball), you can expect Wolfram to make these decisions based upon what makes good business sense. I'd like to respond to one other point in your email, too. It is definitely *not* true that Wolfram has a homogenous development platform. There are numerous Linux and Mac boxes on developers desks throughout the company. I happen to think it's a great advantage that Wolfram has that we don't have a "porting" department because Mathematica is actively developed on every platform we ship (even on Solaris...although at a much smaller percentage than the other platforms). And many developers have two or three computers on their desks, so they can constantly develop and test on multiple platforms. However, it does tend to be true that more talks are given on Windows than other platforms. This is just a reflection of the fact that, for live talks, Windows and Mac machines tend to be better for presentations and that, for screencasts, we have a standardized Windows recording setup. Sincerely, John Fultz jfultz at wolfram.com User Interface Group Wolfram Research, Inc.