Re: use of ATLAS library in octave

[Himanshu B. Dave]

Precompiled binaries have their own problems. For one, they are tied
to a particular distribution.

One can not easily change the installed distribution A, just because to
use some package B, distribution C has to be used. 

Also, most of the distributions like Debian, Suse, Red-hat have trimmed the
code (and more importantly  kernel)  which are away from  standard linux kernels.
This can create  a lot of problems, if the package you are trying to install is using
a different library or unknown library.

This is a problem, to be accepted as such and dealt with. 
I am sure octave team would like it to be used on as many version of Linux, other
Unix like systems and  :( MS-Windows. By saying that "if you want to use octave
get a binary distribution" are we not going some-what away from the spirit of OSS
and Free Software? If I am going to use octave or any other software for a serious
work, I would like to be able to look into the source C/C++ or whatever it is.

John has said somewhere "it is tricky if you do not know what you are doing."
Please allow me some amount of knowledge and skill, otherwise I would have
dared to install octave from  source :-)) I routinely write device-drivers and kernel 
patches as part of my work. Words like PCA, SVD, FFT and wavelet 
transforms are not greek to me. 
As for the "confusion" it was started because of a wrong library in the octave
distribution. I am not blaming any one, in fact I am acutely aware of the tremendous
amount of work that has gone into octave. Only, all of us would like to see that octave 
is available easily to widest range of potential usrs, wouldn't we? User of Binary
Distribtion  AND blessed souls like myself, who dare to install from source!

I very humbly asked to be excused if I have ruffled some feathers.

On 10/25/07, Quentin Spencer <address@hidden > wrote:

John W. Eaton wrote:
> On 25-Oct-2007, Himanshu B. Dave wrote:
>
> | First let me say that octave is an excellent effort at making available
> | a open-source software comparable to a well-known commercial one.
> | I would like to see that more and more enineering students, researchers and
> | engnineers use it. But,  for that to happen, the installation of octave
> | should
> | be as painless as possible. For example, a mechanical engineer,
> | enthusiastically
> | trying to install octave, may find considerable hiccups.
>
> Yes, it has become somewhat tricky to install Octave from sources if
> you don't know what you are doing.  I don't know how to avoid that.
> It's a complicated system with many dependencies that tries to work on
> a wide variety of systems.  But in any case, there are binary packages
> for most systems that take care of all these details.  Ordinary users
> are encouraged to use the binary packages, not build Octave from
> scratch.
>
> | I have come across one. This is due to the developers of octave using an
> | internal
> | function of a published library (ATLAS). This is NOT good programming
> | practice.
>
> You've misunderstood the situation.  As I explained in a previous
> message, Octave does not call dormrz directly.
>
> | You are not supposed to use internal, unpublished functions of a library.
>
> You are confused about what is happening and now you seem to be
> implying that we are sloppy programmers.
>

Compilation of octave is fairly involved, but installation is very easy
if you use a precompiled binary. There are binaries available for
Windows, Mac, and many Linux distributions. Did you ever try using
precompiled binaries?

Quentin