--- Begin Message ---
|
Subject: |
Suggestions for Improving Octave |
|
Date: |
Tue, 31 Jul 2001 21:22:20 EDT |
Hello Developers,
I am not requesting to be put on the help-octave mailing list, I just wanted to
provide some feedback for those of you who are developing Octave and/or Octave
MPI and are looking to improve the software and installation process.
This past summer semester, I had a project for my Cluster Computing Class at
UCF which involved installing Octave MPI on the school's cluster and running
various tests to see when the number of processors used peaked in terms of
efficiency as compared to the network traffic over the cluster's network.
Unfortunately, I never was able to install Octave MPI and I would like to share
with the Octave community the problems that I had installing both Octave and
Octave MPI and offer suggestions to aid students like myself that concentrate
more on the use of Octave than on the development of it (though I find the
development interesting, that was not the point of my assignment).
1. There are many forms of documentation for Octave, could the documentation be
summed into one file rather than many?
For instance, there is an INSTALL and an INSTALL.OCTAVE file, both of which are
very similar, at least at first glance. But the INSTALL.OCTAVE file introduces
a list of command-line arguements where as INSTALL is not that detailed. But,
INSTALL.OCTAVE has an error in a critical line that is not in INSTALL
(INSTALL.OCTAVE gives an incomplete flag listing when it says to type "make
(FLAGS)" as compared the same statement in the INSTALL file.) So it is unclear
at times when to use INSTALL and when to use INSTALL.OCTAVE. Now, aside from
those to files, there are other installation files (i.e. README.mpi, README,
...) which are either easy to overlook, or conflict or are not as complete with
information in other files (like the INSTALL.OCTAVE or INSTALL file). I guess
what would reduce the confusion is to compress the help files into one help
file. And for OCTAVE-MPI, only include a complete README.mpi document in the
documentation so the non-MPI option is not taken first acc!
identally. Also, see the next s
uggestion, #2, for confusion with missing autoconf directive in INSTALL.OCTAVE.
2. Step 5. of INSTALL should explain that autoconf is not an executable file
that should be included with the package. I was under the impression, from the
way that it was worded, that 'autoconf' was an executable file that if it
appeared in the root of the build tree when I unzip OCTAVE, then I should run
it. That entire Step 5 can be reduced to: "Type 'autoconf' in the root of the
build tree." The same should be true for Step 2. a. in README.mpi. The
'autoconf' step does not even appear in INSTALL.OCTAVE which confused me more.
3. No matter the size, I think there should be a downloadable executable for
OCTAVE-MPI. There really is no need for me to compile the software. There are
too many variables that are in the mix - different versions of compilers,
different versions of MPI, etc.
4. If it is intended that OCTAVE-MPI should only be built on a system that
includes LAM/MPI, then it should be said more clearly in the documentation. In
fact, maybe it should just be a requirement to build and run the software. Or,
suggest to the user in the documentation that they may need to change the -lmpi
links in the configure file to the appropriate links for their version of MPI.
5. The configure file could use a little user friendliness. The '--enable-mpi'
option, for example, should be a prompted question when running the
configuration file than a mysterious command-line arguement which is obvious
anyway and should be a default since I downloaded OCTAVE-MPI because it has MPI
support. There are more opportunties for similar prompts like for 'prefix' and
'location of MPI links' and so on.
6. The help files online at http://www.octave.org/archive.html are catagorized
by year and date. That would be great if I knew when every installation
problem could have happened over the past nine years, but I don't. Also, some
of the titles are misleading or wrong for the help, so I pretty much have to
read every help file. I would appreciate if you can reduce my O(n) search to
an O(log n) search by alphabetizing the topics by title and editing the title
so that it accurately reflects the issue and solution. Maybe in the future, a
search engine type utility could be used to search the archives.
7. I think the idea that people, on their own time, contribute to the OCTAVE
cause for free is a wonderful thing. But with that in mind, since people are
writing source code that will be edited by other people, some common practices
should be enforced like comments in code, list your name and change to the file
(maybe e-mail if that person doesn't mind), etc. After all, free or not, it is
your work and reflects on you so you should take credit for positive changes
and likewise take credit for problems that may arise from a change.
I hope my ideas were constructive and helpful. I can tell that OCTAVE is a
very powerful tool from what I have read online and seen in some of the bug
reports. But installing OCTAVE is the most critical thing, and if that is not
perfected than no one will ever be able to see what kind of power it really has.
Thanks,
Danny Lacks
Grad Student at the University of Central Florida
PS
CC sent to my instructor
--- End Message ---