Proposal to reduce anti-bundling requirements

[Ian Malone]

On 14 September 2015 at 16:47, Jóhann B. Guðmundsson <johannbg at gmail.com> wrote:

> They simply have welcomed their new container overlords and are using only
> the recommended upstream method for installing for their application (
> pip,gem etc since developers can use the upstream support community for
> those ) in those container images, followed by a strong attitude of use it (
> their produced container/vm image not some downstream shipped/provided
> "package" ) or "take your freedom of choice and get lost, we are done trying
> to support and play the "X-factor of linux distributions" game. "
>
> And once the "market" has ( started ) taken a stance ( moving away from
> downstream provided package of their components since it does not work due
> to the fragmentation in the linux ecosystem ) it's irrelevant what I think
> or you think or distributions think or implement locally beside providing
> the underlying structure to run their application for the sole purpose of
> being relevant as an platform for deployment ( which today is basically any
> distribution that ships systemd ).
>

Ultimately that is going to be self limiting, you can only do it while
the most fundamental components are playing by the old rules. I can
think of a few research software packages (in the sense of software
packages, not fedora packages) which are so tied to particular
underlying libraries that getting them to work in the same environment
is a real pain (various ones that bundle underlying libraries and have
their own setups that force that on the whole system because they
can't even get linking right). Now you can containerise that, but
eventually you are going to have to have containers within containers,
and somewhere in there will be a piece of rotting software.

-- 
imalone
http://ibmalone.blogspot.co.uk