[Bio-packaging] Packaging

[Pjotr Prins]

Since we are all subscribed to this list there must be some interest
in packaging and deploying bioinformatics software. As these things
go, packaging is unsexy and gives NO credit in science. But someone
has to do it and I strongly feel that collaboration is worthwhile.

Let's collaborate, or at least make visible what we are doing
individually.

It would be good if we find some consensus on approaches, so as (for
once) try and prevent reinventing the wheel. I do think we can combine
package systems (as CloudBiolinux effectively does).

I like GNU Guix and Docker because they are Linux distribution
agnostic. They both require Linux, but I don't think that really is a
problem. Deployment for me (at this stage) is mostly about batch
processing on Linux clusters. I run Debian, but I also have to work on
CentOS. Agnostic is good (here). 

Ricardo has been working on GNU Guix packaging for bioinformatics and
he has a nice writeup here:

  http://elephly.net/posts/2015-04-17-gnu-guix.html

Peter has written that he is working on a Docker solution for
Biopython. Maybe he can expand a bit on that. And then we also have linuxbrew 
which is Brad's favorite solution. Brad? 

Linuxbrew is the only one that does not require a daemon (that means
root rights). But I am not satisfied by its dependency handling and
versioning of packages. Also (I think) there is no binary support for
packages.

Anyone think we should look into other alternatives? I have worked
with the Debian Biomed packagers in the past but find Debian packages
archaic, error-prone, too hard and too centered on one distribution. I
never took to it - I don't want things to be unneccessarily difficult
when it comes to software deployment.  That is what Debian packaging
is to me. Also Debian packaging does not resolve dependencies,
reproducibility, nor versioning. RPM is in that same boat, so I have
not even tried that.

Pj.