Re: r-cran-maptools_0.7.16-1_i386.changes REJECTED (fwd)
- To: Roger Bivand <Roger.Bivand@nhh.no>
- Cc: Mark Hymers <ftpmaster@debian.org>, Julia Koschinsky <jkoschin@asu.edu>, debian-science@lists.debian.org
- Subject: Re: r-cran-maptools_0.7.16-1_i386.changes REJECTED (fwd)
- From: Andreas Tille <tillea@rki.de>
- Date: Wed, 11 Mar 2009 11:32:35 +0100 (CET)
- Message-id: <alpine.DEB.2.00.0903111122400.15302@wr-linux02>
- In-reply-to: <alpine.LRH.2.00.0903110848341.4911@reclus.nhh.no>
- References: <alpine.DEB.2.00.0903101359110.1963@wr-linux02> <alpine.LRH.2.00.0903101809580.2516@reclus.nhh.no> <alpine.DEB.2.00.0903102243550.15332@wr-linux02> <alpine.LRH.2.00.0903110848341.4911@reclus.nhh.no>
On Wed, 11 Mar 2009, Roger Bivand wrote:
Yes, you miss the main point of examples in R packages. They serve two
functions, one to demonstrate the working of the function on real data (here
data sets used most often in the scientific literature), for import functions
these must be real external files, not saved R objects. Secondly and most
crucially, R CMD check <pkg> is the key QA tool for packages, and runs all
the examples in a package. Any failure in these shows that an edit had
unexpected side effects. I often work on packages offline, so the package
must ship with the real files.
I'm not fully convinced because a "source representation of the data", a
recipe how to build the ESRI Shapefiles and a MD5 sum would do the same
trick - but for simplicity reasons I understand your point.
You can ask the admin at ASU, but I guess that their description is like CC
BY ND, and I don't think you'll get any more there (CC'ed, attached first
email from Andreas with copy of Mark's rejection for Julia's information- I
guess CC BY SA is Mark's minimum requirement?).
Thanks for forewarding the question.
Then let them use another OS and distribution if Debian can't manage for
reasons of its own choosing. This all works for Task Views, and indeed your
time and effort would be much better spent on contributing an Epidemiology
Task View to CRAN. Then any user on any platform could do this automatically,
right?
If you restrict epidemiology to R software yes, but in general no and
there is more software out there which is not using R. Chances to
integrate this into Debian which is done by the Debian Med project are
good. I see no reason to discuss the choice of a distribution at this
point.
Wrong. Running the examples, and especially examples using the sids (North
Carolina sudden infant death syndrome) data set may be crucial to
understanding how to use their own external data, also for epidemiologists.
Contributed packages are both software and domain knowledge, and examples are
crucial to learning.
ACK.
I've CCed Julia at ASU, but their conditions for (many) data sets made
available by (many) researchers are unchanged over many years.
Thanks for the CC anyway.
Note that nothing in this very short document discusses this case. I am very
sure that the same situation affects the vast majority of R contributed
packages that include data sets (especially external file format examples),
but here zealot Mark has put his foot down only because I was careful to
actually bother to write a LICENSE file. In most other cases, things slip by
unnoticed.
You probably have a point here. Thanks for your careful work on the
LICENSE file.
With regard to code and documentation, I have no objections, but I
do have objections wrt. key example data sets, which do not need to be
licensed in the same way.
This has to be discussed inside the Debian project. Thanks for the hint.
Kind regards
Andreas.
--
http://fam-tille.de
Reply to: