Outdated links to the Bioconductor repos in R.app

On Wed, 24 May 2006 Herve Pages wrote:

Quoting Simon Urbanek <simon.urbanek at r-project.org>:

On May 23, 2006, at 10:31 PM, Herve Pages wrote:

[SNIP]

but they should be (in this order):

   http://bioconductor.org/packages/bioc
   http://bioconductor.org/packages/data/annotation
   http://bioconductor.org/packages/data/experiment
   http://bioconductor.org/packages/omegahat
   http://bioconductor.org/packages/lindsey
   http://cran.fhcrc.org

Should CRAN be really on that list?

Well it depends what you are trying to achieve: if you want the GUI
to show the user the list of Bioconductor packages only, then you
don't need the 3 last repos.
But if you also want the user to be able to install from this list  
and
then to load the newly installed stuff, then you need 2 things:

  1) provide a set of repos that is closed for the dependency  
relationship

  2) provide the user with the option to have all the dependencies
     installed in addition to the packages s/he selected

Since the GUI doesn't support 2), then I agree that we don't really
need the 3 last repos (omegahat, lindsey and CRAN)... But I'd rather
keep them anyway because:

  - Some users might want to do things like:
      install.packages('pkg', repos=getOption("BioC.Repos"),  
dep=TRUE)
    The only way to guarantee that this will work for _any_  
Bioconductor
    package is to have all 5 repos returned by getOption 
("BioC.Repos")

Forgive, but if that's the case, why can't it be specified as such:

whole.enchilda.repos <- c(getOption("BioC.Repos"),
                         "http://bioconductor.org/packages/omegahat",
                         "http://cran.fhcrc.org")  # getOption 
("repos")?
install.packages('pkg', repos = whole.enchilda.repos, depend = TRUE)

  - Things are consistent with the biocLite() function that we  
provide
    for users of _all_ platforms
      http://bioconductor.org/docs/install-howto.html

Since the above would be an internal change to the install
script, the above would be even less effort.

The inconvenient of this is that we end up with the GUI
displaying a very long list... but this is not really a
problem because I don't think many users actually browse
the entire list: they'd rather use the search box instead.

I don't. There may be others as well... I'd rather BioC
packages be kept separate from CRAN packages in the
listing...

----------------------------------------------------------
SIGSIG -- signature too long (core dumped)