Re: Re: incompatibilities between releases of Mathematica (was: Mathematica Link for Excel and Excel 2000) Organization: Princeton University - CIT/IS/ASIG
- To: mathgroup at smc.vnet.net
- Subject: [mg19511] Re: [mg19376] Re: incompatibilities between releases of Mathematica (was: Mathematica Link for Excel and Excel 2000) Organization: Princeton University - CIT/IS/ASIG
- From: weber at math.uni-bonn.de (Matthias Weber)
- Date: Sat, 28 Aug 1999 15:53:27 -0400
- Organization: RHRZ - University of Bonn (Germany)
- References: <7pt5o8$ndu$2@dragonfly.wolfram.com>
- Sender: owner-wri-mathgroup at wolfram.com
In article <7pt5o8$ndu$2 at dragonfly.wolfram.com>, David Withoff
<withoff at wolfram.com> wrote:
> > "P.J. Hinton" wrote:
> > >
> > > Perhaps you could help us out by describing in precise terms exactly what
> > > you mean by "broken." The likelihood that you will get your code working
> > > to your satisfcation will increase if you can post specific, minimal
> > > examples that reproduce the problems you've been running into.
> >
> > Actually, I'm not talking about custom code, although that would also
> > be affected by the changes in v4. You can just see by the latest addition
> > to the WRI website regarding free updates to application packages that v4
> > is *NOT* backwardly compatible. Lots of notebooks that worked in v3 also
> > are "broken."
> >
> > Other software packages take pains to insure backwards compatibility, MS
> > Office, etc, so that you can read documents written in prior versions.
> > When a class or method is deprecated in Java, you can still use it with
> > warning flags generated so that you can update the code at a later time.
> > There is a gradual process of "improvement" instead of things just breaking.
> >
> > Now, I expect more from WRI because I have a higher opinion of the company
> > than let's say Microsoft. It would be helpful to have, instead of just a
> > list of imcompatibilites between versions, an explanation of why those
> > changes were made, wordarounds for them, etc. Maybe they are there in the
> > documentation and I just haven't found them yet. I'm all for progress and
> > enhancements but been a programmer for 20 years, I know the importance of
> > legacy systems. We may like to get rid of them but they are going to be
> > with us for a long time. And programmers hate going back to fix legacy code.
> >
> > Joe Yoon
>
> I believe your criticism about backwards compatibility is misdirected,
> and I second P.J. Hinton's suggestion that you provide specific examples.
> A single example is more likely to be an anomaly rather than a foundation
> for broad criticism, but at least it would be a start.
>
> All of the true backwards incompatibilities that I know about with
> Version 4 of Mathematica are listed in Appendix A.13.4 of The Mathematica
> Book, and none of them have led to much trouble.
>
> All of the changes that have led to trouble are things like errors
> in Mathematica, and undocumented behaviors that were changed or removed
> and that some people were apparently using in their programs. It seems
> misleading to describe such things as backwards incompatibilities.
> If someone exploits an error in Version 3, for example, it doesn't seem
> quite right to describe it as backwards incompatibility if that persons
> program breaks because Wolfram Research fixed the error.
>
> If you *do* have an example of a significant backwards incompatibility
> that isn't already documented, it would be useful if you could report
> it so that we can work to avoid such problems in the future.
>
> Dave Withoff
> Wolfram Research
Below is an annoying MathSource example which I tried to fix
without success, because the package contained code from Mathematica 1.2 which was
abandoned in 2.2 and later -- like Flatten being called with negative
levels or expressions like Det[ {IdentityMatrix[3],{1,2,3},{4,5,6}}].
I had a little chat with the suppor about it without getting much help.
Apparently, undocumented features of these functions were used in the package.
I don't have a version of Mathematica around which is that old, so I couldn't
even check whether the package worked once.
It would be nice to have packages like BackToMathematica 1.2 which re-implement
older versions of Mathematica so that one can at least check whether a
certain (mis-) bahaviour of a package in a new version is due to the new
version or something else.
Anyway, if somebody is interested in maintaining legacy code:
http://www.mathsource.com/Content/Enhancements/Geometry/0205-108
Matthias