[ckan-discuss] sudo apt-get install data-pkg

Mon Sep 12 07:54:20 BST 2011

Hi Félix

...data packaging will make easy for the non-technician download,

Sure, but for the non-technician even asking them to use command line may be
too much, no? I'd argue (and I think you agree) that websites or GUIs would
be the best solution for them. Then, again, debian packaging would allow
people to just use Synaptic (or another apt GUI) to "install" data just like
software packages.

you must look after non debian-like linux users.

You're absolutely right. I did not suggest to replace datapkg, I was rather
suggesting an extra option for those that already package their software as
deb.

Maybe could be a good idea to provide some methods on datapkg so they
> could auto-generate a debian package from the data package source

That sounds like a great idea. I don't, however, have much experience with
python eggs or debian packaging, so I'm not the best person to try this. For
my use case and available time, the hybrid solution is probably the best.

Thanks again for your contributions to the discussion.

Cheers,
Pablo

2011/9/12 J. Félix Ontañón <fontanon at emergya.es>

> El día 6 de septiembre de 2011 10:45, Pablo Mendes
> <pablomendes at gmail.com> escribió:
> > Hi Félix,
> > Thank you for your opinion. This solution may do exactly what I need with
> > some postinst hooks. However, I disagree with the radical distinction you
> > seem to make between data and programs. They definitelly have some
> > particularities wrt the storage/download/install process, but on the
> surface
> > they are just packages that depend on each other.
> >
> > *.doc files depend on Libre/MS Office to be read
> > Powerpoint/Impress depend on a database of cliparts
> >
> > (just to cite day-to-day examples)
> >
> > The current solution is to have the dependencies between data and
> software
> > (and vice versa) on the hands of the user, often providing custom
> software
> > to help her/him with the install process, but many times making him/her
> > download files and place them in some requested folder.
> >
> > For me, this would be a better world if I could use just one installer
> that
> > took care of dependencies between all kinds of packages (software and
> data).
> >
> > Speculation: maybe the low uptake on datapkg is due exactly to this clear
> > cut distinction. If you are requested to download a file during the
> > installation process, it should be easy enough to just open the browser
> or
> > use wget. Using datapkg would in fact require two steps: first get
> datapkg,
> > then get the file you want. However, if the data download could be
> triggered
> > through the software install, then you wouldn't need to ask the user for
> > anything.
> >
> > This is what I will try: hide the usage of datapkg inside my deb package.
> >
> > Cheers
> > Pablo
>
> Hi again Pablo,
>
> As i've learnt, okfn people also try that data packaging will make
> easy for the non-technician download, update and upload packages to
> ckan too.
> Am I wrong?
>
> Although I think you're right about why to distinct data and software,
> you must look after non debian-like linux users.
> The success of python-egg or java-jar is to provide a way to
> distribute, configure and deploy software for every platform.
>
> Maybe could be a good idea to provide some methods on datapkg so they
> could auto-generate a debian package from the data package source,
> just like that methods on python distutils which can be use to
> auto-debian-package a python egg.
>
> Kind regards,
>
> > On Sep 6, 2011 9:11 AM, "J. Félix Ontañón" <fontanon at emergya.es> wrote:
> >
>
>
>
> --
> J. Félix Ontañón Carmona
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.okfn.org/pipermail/ckan-discuss/attachments/20110912/4ce796a0/attachment.htm>