Re: informix upward compatibility (fwd)
In <1992Aug11.163009.4870@dlss2> james@dlss2 (James Cummings) writes: >When we got down to inquiring about the upgrade we find that >the upgrade would not be guaranteed unless our OS was the OS that the >upgrade was ported to. We have a 3b2/700 that is currently on SysV r 3.2.1 >the Informix upgrade was ported to SysV r 3.2.3 (hmmmm...). "What's the >*real* difference?", we asked...Informix just will not stand by the upgrade >unless we were on the right release was the only reply. Additionally, there >should be another higher release of SQL, but it was ported for SysV r 4.0! >Informix wanted to know why we didn't just upgrade to rel 4 of SysV and get >their latest/greatest. Folks, on a 3b2 that a major HARDWARE & SOFTWARE >upgrade costing large denomination dollar bills. I suppose what is >troubling to me about this is, that Informix addresses their "known" bugs >in upgrades ported to the current release of the OS. That's well and good >if "everyone" is going to "always" upgrade the OS. But not to support >their product except in this way seems to abandon the end user to and >upgrade everything or die existence. I was in the business of supporting a Unix package on many hardware platforms, and I understand very well the Informix position. Their current distribution problem is a three dimensional matrix: one axis is the hardware platform, another axis is the operating system version, and the third is the Informix software version. (I won't even talk about the distribution media part of the matrix). Each port to a platform involves a unique combination of these three factors. From your description, Informix has distributed, just for the 3B2, the following combinations: Hardware/ 3B2 3B2 3B2 Op Sys V.3.2.1 V.3.2.3 V.4 Informix 1.03 4.00 4.10 Maybe I got the details wrong, but that is the general idea. Now, you want them to be more flexible. For this one hardware platform, at this point in time we have three operating system versions and three Informix versions. (I suspect there are both more Informix versions and Unix versions on the 3B2). If they have to support every combination, then rather than three products (ports), they would have to produce 9 different ports. And just for this one platform. How many platforms do they support? 20? 30? Do you really expect them to triple or quadruple their porting work just to support obsolete hardware or operating system platforms? Have you ever ported a large body of software to multiple platforms? Let's assume that the code always compiles, and no work is involved there. Consider just the logistics. It is something like this: 1. Load target operating system (Remember we are supporting multiple operating systems). 2. Load source. 3. Compile source. 4. Run regression tests to test software. 5. Build distribution media. 6. Remove all source from machine. 7. Load distribution disks (test the distribution!) 8. Run regression tests again. 9. Repeat for another combination of operating system and vendor software. You wanted a *guarantee* from Informix that your bug was fixed. For them to give you a guarantee, they must have tested the distribution on a particular hardware/software/Informix combination. They do not *know* if it will work on other combinations. It might, but they cannot guarantee it. Informix certainly has no control over bugs introduced, say, into the C runtime library, and are discovered when you relink new Informix runners. The possibility for errors that are out of their control are bounded only by the imagination of other software suppliers. BTW, it is not Informix's fault that AT&T charges lots of money for upgrades. Sorry for the diatribe, but I have been there, and I know how nasty a problem it is to support *open systems* that are all *compatible*. -- ----------------------------------------------------------------------------- | Marlin Prowell | There is a very thin line between ignorance | | (206) 676-1554 | and arrogance and I have totally obliterated | | mbp@nyssa.wa7ipx.ampr.org | that line. -- Dr. Science |