Comments from ftpmaster regarding spdep Debian package

[Roger Bivand]

On Wed, 30 Jul 2014, Andreas Tille wrote:

> Hi Roger,
>
> On Wed, Jul 30, 2014 at 09:36:22PM +0200, Roger Bivand wrote:
>> Hi Andreas,
>>
>> This isn't a patch to my spdep/inst/README file.
>
> No, it's not! :-)
>
>> I have no knowledge
>> of any debian/copyright file, and cannot take responsibility for
>> that
>
> This was never intended - it was just for our ftpmasters because its
> their responsibility to verify the copyright + license of each file we
> want to upload.  Sorry if this was confusing.
>
>> (I don't think installing from source is a problem for
>> non-OSX/Windows users; I do not use Debian systems, and do not know
>> anything about their packaging systems other than very bad
>> experiences with people with messed up GIS packages). If you want me
>> to patch anything, diff from spdep/inst/README on R-forge. In
>> particular:
>>
>>  Copyright: 2005 Yongwan Chun, Michael Tiefelsdorf and Roger Bivand
>>  License: GPL-2+
>>
>> looks very wrong.
>
> Well, the line above this was
>
>  Files: R/SpatialFiltering.R
>
> in front of the Copyright/License paragraph.  We need to list all
> explicite Copyright statement inside the code.  I can not see in how far
> this should be in conflict with your spdep/DESCRIPTION file.
>

Please understand that the by-file copyrights are not important in R 
packages that I have written and maintain. The DESCRIPTION file is the 
root definition. I'm not prepared to check every file in spdep (or other 
packages) for copyright definitions, as for R purposes these are covered 
globally. If Debian "need to list all explicit Copyright statement inside 
the code", that is a policy choice that is non-conformant with R packaging 
practice. The dates are almost all wrong, and the names are often wrong. 
There are more details in the ChangeLog, but even they are not consistent. 
Often changes are made based on user wishes on our mailing list without 
clear attribution by person (idea by NN, code changes by me). Sometimes 
there are comments, but not always.

R/SpatialFiltering.R was written by Chun, Tiefelsdorf and myself, and has 
been modified by me since 2005.

>> The list of contributors is in spdep/DESCRIPTION
>> in R standard parsable form;
>>
>> +Files: R/bptest.sarlm.R
>> +Copyright: 1998 Joseph O'Rourke <orourke at cs.smith.edu>
>>
>> isn't the correct file - should be src/soigraph.c.
>
> Uhhmmm, perhaps I misinterpretet your first answer.  We have to
> troublesome files.  Would you please consider reading the original mail
> here:
>

No.

>   http://lists.alioth.debian.org/pipermail/debian-science-maintainers/2014-July/026470.html
>
>> I'd be grateful if you'd run any (unneeded) copyright file by me
>> before proceeding.
>
> I hope not to do any unneeded work.  The Debian copyright file is
> required for any Debian package and you can have a look here:
>
>   http://anonscm.debian.org/viewvc/debian-science/packages/R/r-cran-spdep/trunk/debian/copyright?view=markup
>
> I'd be more than happy if you could clarify the open issues for the
> two files
>
>   src/soigraph.c

This is the problematic file which I covered in my earlier reply.

>   R/bptest.sarlm.R

The code is from a GPL 2 | 3 R package (when copied only GPL2, 
subsequently 2 | 3 (but not 2+, either 2 or 3, not any beyond 3 if they 
ever emerge).

>
> in a format like this.

No way. The Debian formats are no help to me in what I do. If you need 
that format, you should create it.

>
>> If such a file is needed, it should only point to
>> the correct file within the R package (otherwise they will get out
>> of sync).
>
> Since we are packaging more than R packages in Debian we can not drop
> this general requirement and we need to stick to the given format.  I
> agree that keeping these files in sync is a bit troublesome but I have
> no choice but providing such a file (which works for >500 R packages
> in Debian).
>
>> As you can see, the "decision" of a "master" is of very little use
>
> s/master/ftpmaster/ = the gate keeper of the Debian package pool.
>
>> to me, I'll humour your attempts if you do things right, but have no
>> need to see spdep distributed in this way - Debian users should IMO
>> always install R packages from source to avoid unintended
>> incompatibilities.
>
> The rationale why I intend to package spdep is that we have a certain
> set of R packages packaged for Debian for very good reasons and now it
> turned out that spdep is used as a new dependency to run a test suite of
> some other packages.

So this is a dependency problem generated by some other R package that 
imports from or depends on spdep? That is, spatial data analysis is 
tangential to your needs? That makes me even less inclined to help. Which 
package? My guess is stargazer, so the easy solution is to tell stargazer 
to de-list spdep (the model output from models fitted in spdep is totally 
misunderstood and mangled by stargazer omitting the spatial coefficoents - 
I told the stargazer maintainer about this long ago, and asked him to fix 
it, which he didn't).

Unless you explain why I should spend any more time on this, it isn't 
going anywhere, and you are wasting time on spdep. I don't see any 
benefits for my (many) users.

Roger

> To make sure we will not have any
> incompatibilities we really want to run the test suite - thus I need to
> provide a spdep Debian package.  I'm aware that there are people who
> consider Debian packages of R packages useless but as you probably also
> know there are varying opinions about this and I'm just doing it as
> a service request of some users (and not because I have no better clue
> how to spend my spare time ;-)).
>
>> Best wishes,
>
> Same to you and sorry for the confusion I might have created
>
>       Andreas.
>
>

-- 
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; fax +47 55 95 91 00
e-mail: Roger.Bivand at nhh.no