[ESS] A debate on progress: WAS build problem

[Sparapani, Rodney]

Hi Dirk and ESS-help:

It is a slow nice day on ESS-help so I will entertain a debate.
Nice to see that my old friend Dirk has not frozen solid.
The thermometer has recently explored some rare territory.
However, I have to respectfully disagree.  I didn�t feel such a
commitment to progress when I wanted to add multi-way
arrays to Rcpp
https://www.mail-archive.com/search?l=rcpp-devel@lists.r-forge.r-project.org&q=subject:%22%5C%5BRcpp%5C-devel%5C%5D+3D+array+proof+of+concept%22&o=newest

RcppArmadillo seems like overkill here when this can be
accomplished in relatively few lines of Rcpp (and even
fewer if Rcpp supported it).  But, I don�t want to fight
with Dirk over Rcpp.  I can honestly say that Rcpp has had
a bigger impact on my career than ESS so I treasure it
even if I don�t always agree.  Beyond that I consider
myself a data scientist, not a computer scientist.  So,
I leave it up to Dirk and Rcpp core to decide what is best
(they are far more likely to be right than me).

Yet, I definitely part ways with the recent Emacs upgrades.
As far as I can tell, ESS is far more compliant with version
26 than 27 or 28.  Certainly, part of this might be our
fault as ESS developers.  But, I don�t understand why some
of the changes in 27 were necessary.  Why do we have
to set a keydef to nil for some keys before we can give
them a new definition?  This was not a problem for
earlier releases.  Furthermore, presumably, this is not
a bug since the same �feature� appears in 28 as well.
How are we supposed to determine these mysterious
magic keys?  Since it is not evident, we have to unset
any key before assigning it a keydef otherwise it is a no-op.
I don�t see a rationale for this direction.

Also, Emacs has become more difficult to compile
particularly on macOS.  Even 26, is not convenient to
compile with an R friendly tool chain.  27 and 28 are
no better.  So, gone are the days when I could compile
Emacs myself.  Now I am stuck with using binary
distributions which makes debugging more challenging
as well.  Therefore, I see the line of progress leading
to the very stable and friendly 26 has stagnated even
as the release schedule for 27 and 28 has sped up.
So, I have one machine with 26 and one with a
menagerie of 26, 27 and 28 installed for testing
(spurring my commentary).  I use Emacs every day and
I�m not particularly that bothered by this lack of progress.
But, I sincerely hope that we are turning a corner
and 29 will be a better experience for ESS with
no magic keys, etc.

Thanks to all for their support of ESS especially
my fellow developers who give their time so freely!

--
Rodney Sparapani, Associate Professor of Biostatistics
Director for the Wisconsin Chapter of the American Statistical Association
Institute for Health and Equity, Division of Biostatistics
Medical College of Wisconsin, Milwaukee Campus


	[[alternative HTML version deleted]]