Comments from ftpmaster regarding spdep Debian package

[Andreas Tille]

Hi Roger,

On Thu, Jul 31, 2014 at 03:47:34PM +0200, Roger Bivand wrote:
> >I'm very astonished that this thread which started with two very simple
> >questions exagerated to such a time consuming discussion.
> 
> Most questions are not as simple as they appear, think ontologies.
> Mine differs from yours. You haven't toød me why you need to package
> spdep other than that another R package depends on (enhances?)
> spdep.

  http://surveillance.r-forge.r-project.org

> I guessed stargazer, but you ignored my question. I need to
> know why this matters.

I have no idea why you guessed this since stargazer is not packaged for
Debian and I do not know about any intend to package it.  I did not
wanted to avoid your question but rather wanted to save your time with
discussion that seemed irrelevant to me.

> >Since I read
> >a lot of frustration in you mail and no chance to clarify things easily
> >I just repeat the main point: inst/README says:
> 
> A frequently given reply for Debian/Ubuntu users of rgdal (R
> interface to GDAL) is to avoid any and all pre-packaged
> Debian/Ubuntu packages and similar dependency traps. The
> Debian/Ubuntu packaging of central GIS software is so poor and tardy
> that it is a major problem for users.

Since I can not find rgdal in the Debian package pool I would welcome if
this accusation would go to the place where it belongs to.  While I
agree that Debian GIS software -- not GIS software written in R but in
general -- was not in the best state but there is currently an effort to
enhance situation.  It does not help if people repeat third persons
experiences without checking the current status.

> Consequently I view any such
> packaging as unnecessary in principle and practice, and that users
> should install from source on Linux/Unix, especially CRAN packages
> and R itself, and crucial external dependencies.

I'm working in the Debian Med project and the epidemiology task has
several R packages:

   http://blends.debian.org/med/tasks/epi

The packages there do not have any open bug reports and are up to date
with the exception of r-cran-surveillance which I'm working on
currently.  It needs to pass the new queue for the sake of waiting for a
r-cran-maptools.  Since I added auto testing feature for all packages
featuring unit tests also spdep is needed since it is used in the unit
tests.  You seem to be concerned about the quality of packaged software
and so I hope you agree that running a test suite automatically and
periodically is a good idea.

In short:  I accept your opinion about Debian packaging but there are
users out there who are regarding our effort as helpful and I'm doing
this for these people.

> >File: src/soigraph.c is copyright (c) 2001 by Nicholas Lewin-Koh and
> >is subject to the licence at the foot of this file, but also contains
> >subroutines TwoCirclesxx and SubVec, which are adapted for R and Double
> >precision coordinates by Nicholas Lewin-Koh, from Computational Geometry
> >in C, Joseph O.Rourke, Cambridge University Press (1998). Copyright for
> >those subroutines remains his.
> >
> >Can you please confirm that the license is really GPLv2+?
> >
> > [ ] yes
> > [ ] no
> 
> No idea, IANAL, I assume that the original author Nicholas Lewin-Koh
> checked in 2001 when the code came into being that the license given
> was adequate. My interpretation of the text in the file from which
> the functions were taken is that it permits use and redistribution
> provided that the copyright to J. O'Rourke is acknowledged, which it
> is.

I'll forward this to our ftpmaster whether he might regard this as
sufficient to distribute the package.

Kind regards

       Andreas.

-- 
http://fam-tille.de