After a discussion to untangle this confusion I feel better focused - we both had somewhat mis-understandable comments which didn't make it any better :-)
Thanks rbasak to help me clarifying this!
-- Details --
According to popcon [1] (to somewhat guess usage) we look at ~0.27% of the installations actually having the meta package "ubuntu-virt-server" installed.
ubuntu-virt-mgmt is about half of that and the over-arching ubuntu-virt is even fewer.
There is one seed (and an assocated tasksel) as one can see in reverse dependencies [2].
The "Virt-Host" seed [3] is inconsistent - it does not "just" depend on ubuntu-virt or any of its subsets. Instead if depends on ubuntu-virt-server, but then also explicitly libvirt-bin, openssh and various per arch qemu's.
Also these days there is no singular "Ubuntu virtualization stack" anymore. These days virt-stacks are mostly defined by their management interfaces.
If one wants containers he picks LXD. If he is going openstack he will go that route and so on.
Therfore I challenge if it is still important to have a "ubuntu-virt" meta package?
For the tasksel it might makes sense to exist, but even then it should be done via the seed directly.
---
Proposal:
A - drop ubuntu-virt* meta pkg in Artful:
A1. change the virt-host seed to directly on correct binary packages
A2. drop the source package ubuntu-virt in artful (and thereby the meta packages ubuntu-virt, ubuntu-virt-mgmt, ubuntu-virt-server)
Alternative:
B - keep ubuntu-virt* meta pkg, but restructure it's source in >=Artful:
B1. Add seeds for ubuntu-virt-* as we decided we really need them
B2. make the creation of the ubuntu-virt-* meta packages part of ubuntu-meta
B3. drop the direct redundant dependencies from virt-host seed
B4. drop the source package ubuntu-virt in artful
I outlined B right away in case A can't be accepted to quickly get to something actionable.
After a discussion to untangle this confusion I feel better focused - we both had somewhat mis-understandable comments which didn't make it any better :-)
Thanks rbasak to help me clarifying this!
-- Details --
According to popcon [1] (to somewhat guess usage) we look at ~0.27% of the installations actually having the meta package "ubuntu- virt-server" installed.
ubuntu-virt-mgmt is about half of that and the over-arching ubuntu-virt is even fewer.
There is one seed (and an assocated tasksel) as one can see in reverse dependencies [2].
The "Virt-Host" seed [3] is inconsistent - it does not "just" depend on ubuntu-virt or any of its subsets. Instead if depends on ubuntu-virt-server, but then also explicitly libvirt-bin, openssh and various per arch qemu's.
Also these days there is no singular "Ubuntu virtualization stack" anymore. These days virt-stacks are mostly defined by their management interfaces.
If one wants containers he picks LXD. If he is going openstack he will go that route and so on.
Therfore I challenge if it is still important to have a "ubuntu-virt" meta package?
For the tasksel it might makes sense to exist, but even then it should be done via the seed directly.
---
Proposal:
A - drop ubuntu-virt* meta pkg in Artful:
A1. change the virt-host seed to directly on correct binary packages
A2. drop the source package ubuntu-virt in artful (and thereby the meta packages ubuntu-virt, ubuntu-virt-mgmt, ubuntu-virt-server)
Alternative:
B - keep ubuntu-virt* meta pkg, but restructure it's source in >=Artful:
B1. Add seeds for ubuntu-virt-* as we decided we really need them
B2. make the creation of the ubuntu-virt-* meta packages part of ubuntu-meta
B3. drop the direct redundant dependencies from virt-host seed
B4. drop the source package ubuntu-virt in artful
I outlined B right away in case A can't be accepted to quickly get to something actionable.
[1]: http:// popcon. ubuntu. com/by_ inst people. canonical. com/~ubuntu- archive/ germinate- output/ ubuntu. artful/ rdepends/ ubuntu- virt/ubuntu- virt-server people. canonical. com/~ubuntu- archive/ seeds/ubuntu. artful/ virt-host
[2]: http://
[3]: http://