Re: packaging aesthetics

[Paul Kienzle]

On Feb 14, 2007, at 11:57 AM, Søren Hauberg wrote:

> David Bateman skrev:
> > Ok, so if the goal is just to make the package installation of a bunch
> > of m-files easier, then keeping the existing package structure and
> > allowing a bunch of m-files package should be possible. The issue I 
> > see
> > is that we currently have two required files. The first is the
> > DESCRIPTION which is the packaging information and the second COPYING 
> > is
> > the license for the code. Do we relax that requirement? If so, how do 
> > we
> > get the required fields of DESCRIPTION (Name, Version, Date, Author,
> > Maintainer, Title, Description)?
> Name (and possibly version) could come from the file name (e.g. 
> image-1.0.0.tar.gz). The rest are mainly of interest in generating web 
> pages, RPM packages , etc. So perhaps the DESCRIPTION file could just 
> be required for packages in octave-forge?
>
> > The easiest, would be to allow a package with a bunch of m-files and a
> > COPYING and DESCRIPTION file. Is that compromise enough?
> We could remove the need for the COPYING file. I would prefer if we 
> could educate people on the need of licenses, but it's a though fight. 
> Most of the matlab code on the web doesn't have a license, so if we 
> don't require the COPYING file we would appeal to the matlab crowd.
>
> I should note that I'm not a fan of not requiring DESCRIPTION and 
> COPYING, but it would be possible.

DESCRIPTION could be useful for installed packages, for example to 
provide a one-line description of the package when listing all 
installed packages, or an extended description of a particular package. 
 I can also imagine some uses for a programmatically available version 
number.  Requiring a one-line license name would also be convenient.  
The complete text of the license is certainly required for 
redistribution from an archive, but I don't think you want to clutter 
up DESCRIPTION with it, and sticking it in COPYING is as good as any 
other.

Requiring these two files in the package root seems perfectly 
reasonable to me.

Matlab used to do some of this with the contents.m file.  I don't know 
the current practice.

- Paul