[Rcpp-devel] Dependence on GNU make because of $(shell)

On Wed, Nov 17, 2010 at 7:44 AM, Dirk Eddelbuettel <edd at debian.org
<mailto:edd at debian.org>> wrote:


    On 16 November 2010 at 23:28, Dominick Samperi wrote:
    | I explained already how anyone who cares can do a diff and resolve
    | this objectively.

    Please show us such a diff and put some proof into this pudding.

    Or else stop harping about a non-existing issue.  Time to "put up or
    shut up".

    | after Nov 2009, so Rcpp today is a different animal. This thread
    started
    | with your remark that my prior work, work that is the foundation for
    | the current Rcpp package, was left "dead and rotting." The purpose
    | of my reply was to correct this misleading remark.

    Not it wasn't. I will stand by "dead and rotting".

    Look, it's simple. RQuantLib was always a user of Rcpp, and I can
    assure you
    that by late 2008 your code __which had not been touched in 2 years__ no
    longer even compiled under current g++ versions. I was using it. I
    believe
    CRAN had even moved the package off the main page as it didn't
    build, and you
    obviously didn't care for it.  So I fixed that and started making
    extensions;
    see the ChangeLog for the initial changes as well as everything we
    all did
    since. The per-project SVN commit counter for Rcpp is now at over 2400.
    That's 2400 individual changesets, sometimes small and sometimes
    large. In
    the space of two years.  Whereas you left RcppTemplate without single
    character changes in three years when it didn't even build. That's
    what I
    call "dead and rotting".

    And I for one do not think it is a coincidence that you come back
    another
    year later bringing the rot to then _three years_ with a short-lived
    update. And I suspect that without the ongoing Rcpp work you would
    never have
    done that brief camoe re-appearance of RcppTemplate.

    Anyway, "dead and rotting" it was and yes, please do provide proof
    for your
    allegations.


I have already provided proof in the form of your own words Dirk. The
quote from Rcpp 0.8.3 that was cited earlier in this thread first appeared
in Rcpp 0.6.7 (released Nov 8, 2009), shortly after my release of
RcppTemplate 6.1 (release Nov 6, 2009), and before Romain joined
the Rcpp project. Thus if anybody cares to diff, the relevant versions
are Rcpp 0.6.7 and RcppTemplate 6.1 (the name RcppTemplate was
chosen to limit confusion between the package name and the
library name, BTW).

I wonder how the authors of the recently released neural network package
would feel if they saw another package author make similar remarks just
days after the release of their hard work, followed by a wholesale
effort to reimplement their work in another package.

On Rcpp::as and Rcpp::wrap, the first is alternate syntax ("sugar")
for a C++ SEXP constructor, and the function of Rcpp::wrap was performed
by what I called getSEXP(). There was also some use of STL classes
to facilitate streaming C++ to R objects. My versions were in the prototype
phase, not as comprehensive as what was implemented later by
Romain, but the main ideas were there in RcppTemplate.

Shortly after the release of my work others joined the Rcpp team, the
pace of development increased dramatically, and it became clear that
to avoid wasting my time I needed to take my work in a different
direction, so I withdrew RcppTemplate and created cxxPack.

It is ironic that cxxPack is actually just the underlying plumbing for
number of packages that I have developed over the years and was planning
to release to CRAN, but I have been somewhat reluctant to do this
in view of my experiences with Rcpp/cxxPack.

Finally, who decides when a package is "dead and rotting"? The
person who wants to take it over?

Dominick