On Mon, May 13, 2013 at 3:09 PM, Julian Gehring <julian.gehring at="" gmail.com=""> wrote: > Hi Dan, > > >> No, the binaries are not kept. > > > Would it be feasible to keep them? > I'll ask other members of the Bioc core team to weigh in here with their opinions. > I'm thinking of cases in which I want to reproduce the results of someone > else. CRAN stores all versions of older packages in its archive. CRAN also does not have version control, so there is no way to retrieve older packages other than their archive. Note that CRAN's archive only contains source tarballs, not Windows or Mac binaries, so if you want to reproduce results on one of those platforms, you still may require install-time dependencies. Dan > However, > for bioc I would have build packages manually in case they have been > modified later in the same release cycle. > > Best wishes > Julian

Hi Julian the transformation from a package directory checked out from subversion with a previous revision number into a source package is pretty straightforward. It might be nice to have a more comfortable system available, but given the resources available, it is well possible that other construction sites are more important. Since R is not a comprehensive operating system, and relies on many facilities of the host system, I am also not sure whether the idea of reproducing someone's result from their recorded output of sessionInfo in an automated manner is ultimately not a chimera (other opinions welcome :). Eventually, for that purpose you'll have to used the saved virtual machine, as e.g. these guys have done: http://scofield.bx.psu.edu/~dannon/encodevm (Of course, the output of sessionInfo is still extremely valuable for many other reasons.) Best wishes Wolfgang On May 14, 2013, at 12:09 am, Julian Gehring <julian.gehring at="" gmail.com=""> wrote: > Hi Dan, > >> No, the binaries are not kept. > > Would it be feasible to keep them? > > I'm thinking of cases in which I want to reproduce the results of someone else. CRAN stores all versions of older packages in its archive. However, for bioc I would have build packages manually in case they have been modified later in the same release cycle. > > Best wishes > Julian > > _______________________________________________ > Bioconductor mailing list > Bioconductor at r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor