Maxima's bug list is public.

While it is, in some ethical universe, a plus to announce/share bugs,
it is, as a
practical matter, much less useful than one might initially believe. I
don't know
what the FoCM audience looks like, or how thoughtful they are.  Are
they motivated
to use Linux because there is a bug list?  (Is there?)

With a large enough user base consisting of mostly naive users, all
using
a complicated program, a large percentage of reported bugs are simply
not
bugs, but misunderstandings, user errors, or just wrong beliefs of
what
the right answer is. In fact, even expert users sometimes report non-
bugs.

<tangent> I have reported bugs in Microsoft software, and they REFUSED
to accept them,
requiring me first to subscribe to some newsgroup and report it to
other (non-MS people)
who provided such useless suggestions as "reload your operating
system".
I don't know what ever happened to one bug (which was a very specific
issue with text-to-speech processing), since there was an obvious
though clumsy workaround, and I didn't bother to follow it up except
to mention it to someone at MS research.


Every so often people report as a bug in maxima/macsyma that
integrate(1/x,x)
should be log(abs(x)) instead of log(x).  The bug is probably "you
need to know more mathematics."

Furthermore, if you find what you think of as a bug, how likely are
you to be able
to match it against the bug that someone else found?  Assuming that
you have in fact
found a bug, it may have many distinct manifestations.  This is, I
believe, what
Bondarenko's program does. Find one bug and worry it to death.

It is not clear how a user would truly use a bug list.  Perhaps look
at it before starting his day to see if any of the programs he thought
he might use are broken?  (e.g. Oh, I see LU-decomposition may be
broken. Better not use it.)
Or maybe he starts his day by seeing if he can resolve someone else's
bug?  (Generous of him, but probably not the way to add to his CV.)

There are "quality" metrics that you might try to compute based on the
length of time that bug reports linger (etc). Not clear that this is
very useful, though.



On Aug 26, 10:30 pm, William Stein <[email protected]> wrote:
> Hi,
>
> I'm writing a paper associated to my plenary talk at FoCM 2011
> Budapest, and I am tempted to make the following statement: "The bug
> tracking done by the developers of all four of the Ma's is secret;
> none of them publish a list of all known bugs, the status of work on
> them, and how bugs are resolved."
>
> Here Ma's = Magma, Maple, Mathematica, and Matlab.
>

How bugs are resolved???
I think it is fair to say that for each software system, there are
people who look at bug reports and check them out a little (just try
reporting a bug and see what happens).  If they deny there is a bug,
they will tell you;  if they confirm it, they will tell it to some
developer team which will consider how/when/if to fix it.

Sometimes I've gotten reports "fixed in the development system. will
be fixed in next version."

What do you think happens to resolve bugs?

Frankly, I think having someone paid to look at bug reports has some
distinct advantages over posting alleged bugs on a public list.


>  ** Question: Is this actually true?  **
>
> I've always thought it was true, but maybe it isn't.  If anybody knows
> if *any* of the proprietary Ma's make the list of known bugs public,
> can you let me know?
>
> Oh, if anybody knows why they don't (for the Ma's that don't), I'm
> curious.  (I'm not asking for idle speculation, or reasons you make up
> for them not making their lists public.)
>
> I'm going to use the above to segue into mentioning Sage's trac.

If there are bugs from the Sage community that are really bugs (or
alleged bugs) in Maxima, the appearance of a bug report in Sage's trac
is pretty much irrelevant.  If it works the same way with other
subsystems, then it is pretty much irrelevant, unless Maxima
developers are magically informed.

You can, of course, tell people whatever you want, but I think there
are a few fairly heavy-duty users of Maxima who are likely to attend
FoCM. Whether they will say anything or not, who knows.

RJF

> I'll also put a link tohttp://cybertester.com/andhttp://maple.bug-list.org/to 
> emphasize that the Ma's are definitely
> not bug free.

Do you really want to lend credence to a collection of bugs that you
cannot vouch for?
Do you know that these bugs are meaningful?  My impression is that one
can discover an
expression, say E,  such that Maple or Mathematica cannot tell that E
happens to be identically
zero.
Then you can feed some expression involving E into many different
"commands" and have them fail in some way.  Thus a simplification
failure (a class of failures that is inevitable from the
undecidability of zero equivalence), can be mapped into a failure of
integration, series expansion, limits, etc etc.

I assume that not all of VB's bugs are of this nature, but I think
that finding bugs in systems that claim to do "all mathematics" is
like shooting fish in a barrel.

RJF


>
> Thanks!
>    William
>
> --
> William Stein
> Professor of Mathematics
> University of Washingtonhttp://wstein.org

-- 
To post to this group, send an email to [email protected]
To unsubscribe from this group, send an email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Reply via email to