Notes

docker

2014-09-29:

  • Discussion with Dirk on repositories, library paths, versions.
  • library paths: apt-get users the usr/lib path, while user-run install commands (e.g. install.packages) uses usr/local/lib/, path. Dirk recommends that /usr/local/lib/R/site-library is configured to be user-writable for package installation, rather than installing into home.
  • building directly from CRAN
  • building dependencies: apt-get build-dep, needs the corresponding deb-src lines.
  • issues and tweaks to littler see PR #2

2014-10-01:

  • Discussion on minimal images
  • Discussion on analogsea + docker

  • Blog coverage of Dirk’s talk on our Docker work.

  • boot2docker-cli includes linux flavors, so I might get a look at what the docker experience feels like for those poor souls who can only live it through full virutalization. No go on the install methods listed, but the binary seems to work. Unfortunately my laptop cannot run 64 bit virtualbox…

2014-10-02:

2014-09-29:

  • Carsten’s suggestion re docker registeries, is this something that scientific repositories might one day support? (Excerpt from my reply post):

While I see your point that the Docker Hub might not be ideal for all cases, I think the most important attribute of a repository should be longevity. Certainly Docker Hub won’t be around forever, but at this stage with 60 million in it’s latest VC round it’s likely to be more long-lasting than anything that a small organization like rOpenSci would host. It would be great to see existing scientific repositories show an interest in archiving images in this way though, since organizations like DataONE and Dryad already have recognition in the scientific community and better discoverability / search / metadata features. Building of the docker registry technology would of course make a lot more sense than just archiving static binary docker images, which lack both the space saving features and the ease of download / integration that examples like the docker registry have.

I think it would be interesting to bring that discussion to some of the scientific data repositories and see what they say. Of course there’s the chicken & egg problem in that most researchers have never heard of docker. Would be curious what others think of this.

Reading