[Date Index] [Thread Index] [Author Index]
Re: phasing out support for older versions
On 24 Jun 1999, Atul Sharma wrote: > As a licensed user of v 3.01, I was about to order the upgrade to 4.0 when I > read this post. Though I have had nothing but good things to say about > Mathematica technical support and software, the notion of abandoning support > for older versions strikes me as unfair to legitimate users who are getting > along fine with older hardware and software, with no pressing need to > upgrade either. While probably a numerically small group, they purchased the > software in good faith, which isn't being reciprocated here. While I > recognize this is the prevailing paradigm in the software industry, I had > thought WRI was a classier act and am disappointed to read otherwise. > Customer loyalty, afterall, has to be earned. Technical Support has limited the scope of their work to the current and prior major releases of Mathematica for several years now. If a user does ask a question about an older "unsupported" version, that user is not turned away out of hand, but the user is advised that the types of assistance that can be offered is limited and may not be 100 % effective. Whether you're running Mathematica version 1 or 4, the act of evaluating 2+2 hasn't changed at all, but you might run into trouble asking a technician how to connect that Mathematica 4 front end to a DEC OpenVMS-based 2.1 kernel elsewhere on your network. This limitation is a consequence of an ever-evolving computer industry. It is not possible to write software that will be 100 % guaranteed to work on all future releases of operating system/hardware combinations. These are factors that are outside our control. Hardly any platform is immune to this type of change as we can see in these examples. o In the early 90s, Apple switched its Macintosh computers from 680x0 chips to PowerPC chips. It was still possible to run the non-FPU version of some releases of Mathematica 2.2, provided that you disabled the Modern Memory Manager. With the release of MacOS System 7.6 the Modern Memory Manager disabler switch was dropped by Apple, making it impossible to run the old binaries. o In the mid 90s there has been a changeover in binary formats for Linux, namely from a.out to ELF binary format. Most newer Linux distributions no longer support the a.out format, so it is impossible to run Mathematica 2.2 for Linux without an operating system kernel recompile. o The release of Windows NT 3.51 introduced a more secure memory model that rendered the Local MathLink protocol -- developed for 16-bit Windows 3.11 -- unusable. The workaround under Windows NT 3.51 was to use TCP as the communications protocol. However, under Windows NT 4.0, it was no longer possible to use the Microsoft Loopback Adapter to get non-networked machines to use this protocol. o IBM switched its RS/6000 line to 604e CPUs, did not support some older machine instructions that were coded into the Mathematica 2.2 X Window System front end. As a result, newer model RS/6000 machines could not run the X Window-based notebook front end. WRI tries to provide workarounds and patches where possible. Recent Mathematica 3.0 examples include the Apple Appearance Extension patches for MacOS system 8.x and a fix for HP PCL printing problems reported by several Windows 98 users. Maintaing perpetual support for a product may be a noble goal, but hardly any software company has the resources necessary to retain software development staff for the purposes of writing fixes to old code bases on operating systems and hardware that may not be maintainable. -- P.J. Hinton Mathematica Programming Group paulh at wolfram.com Wolfram Research, Inc.