This email serves as an announcement that I plan to swallow up R-devel into the base R package. Why? * It is causing no end of user complaints. The typical R user expects to be able to do a "CPAN" (really, I should say "CRAN") style package install through the R interface: install.packages("RSQLite") This doesn't work unless you have R-devel installed. The average R user is a professor or a student, and neither of them are going to necessarily possess the necessary Linux/Fedora knowledge to be able to understand why this doesn't work like the R documentation says it should. * The size of the R-devel is tiny, about 440K installed. It will not bloat the main package to absorb the .h files and a .pc file into the main package. There are no libraries in the R-devel package. * The primary users of R-devel are R addon packages. They can continue to BuildRequires: R-devel safely (hooray for Provides/Requires). * libRmath (the R shared library) will still be a separate package. libRmath-devel will still be a separate package. As far as I know, nothing in Fedora uses libRmath, nor is libRmath-devel necessary for building R addons. Now, I recognize that this is a violation of the Fedora Packaging Guidelines, and I've given a lot of thought to this over the last few days, and technically, this is a reversal of my earlier stance (which is to stick to the guidelines). However, I think this is a key functionality that Fedora R users expect to just work, and I want to try to make sure that they get the best R experience out of the box that they can. As Martyn Plummer pointed out: There are currently 1533 packages/bundles on CRAN, not counting the recommended packages that are already distributed with R. Of these packages, 420 require to the R headers. Fedora only supplies RPMs for 9 of them. I'm not entirely sure if I need FESCo or FPC approval to take this action, if so, this is my notice of requesting it. ;) I'm also looking for feedback and comments, of course. Thanks, ~spot (Fedora R maintainer)
On Thu, Oct 23, 2008 at 09:55:08AM -0400, Tom spot Callaway wrote:> * It is causing no end of user complaints. The typical R user expects to > be able to do a "CPAN" (really, I should say "CRAN") style package > install through the R interface: > install.packages("RSQLite") > This doesn't work unless you have R-devel installed. The average R user > is a professor or a student, and neither of them are going to > necessarily possess the necessary Linux/Fedora knowledge to be able to > understand why this doesn't work like the R documentation says it > should.Where do R packages installed this way go? I'd rather have people say "hey, we're missing an R package" and then have that packaged up for Fedora, so we can then later do a consistant install via kickstart. (Like perl modules, they're generally relatively easy to package.) -- Matthew Miller <mattdm at mattdm.org> Senior Systems Architect Cyberinfrastructure Labs Computing & Information Technology Harvard School of Engineering & Applied Sciences
Thank you. We really appreciate this. Martyn On Thu, 2008-10-23 at 09:55 -0400, Tom "spot" Callaway wrote:> This email serves as an announcement that I plan to swallow up R-devel > into the base R package. Why? > > * It is causing no end of user complaints. The typical R user expects to > be able to do a "CPAN" (really, I should say "CRAN") style package > install through the R interface: > > install.packages("RSQLite") > > This doesn't work unless you have R-devel installed. The average R user > is a professor or a student, and neither of them are going to > necessarily possess the necessary Linux/Fedora knowledge to be able to > understand why this doesn't work like the R documentation says it > should. > > * The size of the R-devel is tiny, about 440K installed. It will not > bloat the main package to absorb the .h files and a .pc file into the > main package. There are no libraries in the R-devel package. > > * The primary users of R-devel are R addon packages. They can continue > to BuildRequires: R-devel safely (hooray for Provides/Requires). > > * libRmath (the R shared library) will still be a separate package. > libRmath-devel will still be a separate package. As far as I know, > nothing in Fedora uses libRmath, nor is libRmath-devel necessary for > building R addons. > > Now, I recognize that this is a violation of the Fedora Packaging > Guidelines, and I've given a lot of thought to this over the last few > days, and technically, this is a reversal of my earlier stance (which is > to stick to the guidelines). However, I think this is a key > functionality that Fedora R users expect to just work, and I want to try > to make sure that they get the best R experience out of the box that > they can. > > As Martyn Plummer pointed out: > > There are currently 1533 packages/bundles on CRAN, not counting the > recommended packages that are already distributed with R. Of these > packages, 420 require to the R headers. Fedora only supplies RPMs for > 9 > of them. > > I'm not entirely sure if I need FESCo or FPC approval to take this > action, if so, this is my notice of requesting it. ;) > > I'm also looking for feedback and comments, of course. > > Thanks, > > ~spot (Fedora R maintainer) > > _______________________________________________ > R-SIG-Fedora mailing list > R-SIG-Fedora at r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-fedora----------------------------------------------------------------------- This message and its attachments are strictly confidenti...{{dropped:8}}
On Thu, 2008-10-23 at 16:34 +0200, Enrico Scholz wrote:> Better solutions: > > * add it to comps.xml > * move 'R' to R-core, and add 'R' which depends on 'R-core' + > 'R-devel' > > > > * The size of the R-devel is tiny, about 440K installed. > > You miss its dependencies: > > bzip2-devel > gcc-c++ > gcc-gfortran > libICE-devel > libSM-devel > libX11-devel > libXmu-devel > libXt-devel > libjpeg-devel > libpng-devel > ncurses-devel > pkgconfig > readline-devel > tcl-devel > tetex-latex > texinfo > tk-develThese are very good points, thanks Enrico. What would people think about doing the suggested R/R-core/R-devel split instead? Users would still be able to get everything with yum install R, it would meet the guidelines, and minimal installs with R can simply have R-core. ~spot
Tom "spot" Callaway wrote:> These are very good points, thanks Enrico. What would people think about > doing the suggested R/R-core/R-devel split instead? Users would still be > able to get everything with yum install R, it would meet the guidelines, > and minimal installs with R can simply have R-core.Sounds like a good idea to me :) Pierre ~ pingou