MathGroup Archive 2007

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Re: v5.2 preferred for stability over v6.0

On 22 Jun 2007, at 19:38, David Bailey wrote:

> Andrzej Kozlowski wrote:
>> In my case both Mathematica 5.2 and 6.0 give exactly the same answer
>> for this limit (provided you make the necessary assumptions). So this
>> leaves two possibilities.
>> One is that you have a different version of Mathematica than me. That
>> would explain the fact that our experiences are exactly the opposite:
>> I find version 6.0 a giant improvement over 5.2 and I find its
>> documentation center almost exactly what I have always wanted.
>> However, it seems to me highly unlikely that our versions of
>> Mathematica 6.0 are different. In which case the only other
>> explanation is that for your performance as a debugger you deserve a
>> substantial pay cut.
>> Andrzej Kozlowski
>> On 20 Jun 2007, at 18:41, jrc wrote:
>>> With my recent experience with the inability of v6 to
>>> find a simple limit (easily found with 5.2 and, incidentally,
>>> with another system), and reading over the confusion in the posts
>>> here recently, I've decided to do any serious analysis with
>>> v5.2 instead of v6.0.
>>> This seems to be another typical case of *.0 versions full
>>> of buggy new ideas with incomplete development. The worst
>>> part is the hopelessly misconstructed and unexplained new
>>> 'documentation center'. Whatever improvements there are in
>>> v6.0 are so incompletely developed, and poorly explained
>>> (if at all) as to not be worth the upgrade.
>>> Hopefully Wolfram is paying attention, and I would guess
>>> that we are the primary debuggers. I, for one, would like
>>> a pay increase.
>>> jrc
> Andrzej, I don't think those comments are fair. The fact is that  
> people
> build large and complex systems using Mathematica. You don't need many
> tiny changes in the behaviour of Mathematica to create havoc for
> developers. The effort to reduce a problem to something that can be
> presented here can be substantial and nerves can become frayed!
> I myself have spent considerable time recoding certain plots that  
> worked
> fine at 5.2, but have become unbearably slow in 6.0. If they had been
> slow at 5.2, I would have accepted that with a shrug and recoded them
> accordingly, but it is the change that can be very disruptive.
> I am not saying I agree with JRC, but clearly there are changes and  
> new
> bugs with 6.0, and some people's code is inevitably hit harder than
> others on a more or less random basis!
> David Bailey

I can't really see what you think was unfair about my comments, or  
rather, in my opinion, if there was anything unfair than it had  
little to do with the points you make. I confess I made one mistake -  
I did not carry out all the steps that JRC did in trying to compute  
his limit. More precisely, I did  not use $Assumptions for the reason  
that I never use it - in my opinion the global (as opposed to local)  
assumptions mechanism is one of the few rather daft ideas in  
Mathematica. Or perhaps, not really daft, but completely unusable for  
a naturally forgetful person like myself. So when I computed the  
limit that JRC claimed (quite wrongly) Mathematica could not compute  
using local assumptions it returned exactly the same answer as in  
Mathematica 5.2. I then wrongly (or perhaps "unfairly") assumed that  
JRC did something different in computing the Limit when using version  
6 and when using version 5.2. Eventually, as a result of off-line  
communication with him, I realized that there was something wrong  
with $Assumptions.
But does this limit problem and all the comments that JRC made about  
Mathematica 6.0 have to do with "people build large and complex  
systems using Mathematica"?? As far as I can tell now (see the post  
by Daniel Lichtblau) he found a bug in $Assumptions. He then  
complained that Mathematica 6.0 could not compute a simple limit that  
Mathematica 5.2 and another CAS has no problems with. This was  
manifestly false: Mathematica 6.0 has no problems with this limit  
provided the right assumption are passed to Limit.  Later JRC  
complained  that the new behaviour was not documented. Since it is  
the result of a bug, how on earth would you expect it to be documented?
And then finally, what does this limit problem have to do with all  
the things in your post? JRC suggested that Mathematica 6.0 was  
either not worth the upgrade price or (depending on how you interpret  
his comments) it was not sufficiently debugged before being released  
and people like himself have to play the role of debuggers. You seem  
to suggest that you do not agree with  him (?) but then what exactly  
do you agree with? I wrote, ironically, that as a debugger he was not  
doing a very good job - he claimed that mathematica could not compute  
a limit which is certainly could easily compute. I also don't agree  
with a single one of his comments about Mathematica 6.0, in fact I  
think they were the most unfair comments about a new release of any  
program I have read in a long time. If you now think this is unfair  
than perhaps you should explain more clearly what exactly you think  
about these claim. I don't see how saying that "some people's code is  
inevitably hit harder than others on a more or less random basis" is  
relevant. Isn't that true about any major new release? Are you saying  
the release of v. 6 should have been delayed for another half a year  
or perhaps longer?

Andrzej Kozlowski

  • Prev by Date: Re: Re: My problem when solving a system of equations
  • Next by Date: Re: Solving a Integral (2)
  • Previous by thread: Re: v5.2 preferred for stability over v6.0
  • Next by thread: Re: v5.2 preferred for stability over v6.0