Thanks for the feedback.

On Fri, Sep 22, 2017 at 05:46:36PM -0000, bp wrote:
> You can skip this paragraph if you're in a hurry.  First, thank you for
> holding this discussion in the open.  I totally understand the marketing
> reasons for this: I depend on Canonical's distribution, and want Canonical
> to offer their paying customers a fantastic out-of-the-box experience so
> that they can continue to offer their services to the general public, too.

I want to be clear that, although Canonical benefits financially from users
signing up for paid extended support through ESM, the rationale for
including the package in ubuntu-minimal is not a "marketing" one.  Rather,
the intent is to ensure users who have opted into the paid ESM service have
a seamless process for enabling it on their system.  We do want users whose
systems are EOL and no longer security supported to know what their options
are (upgrades, or paid ESM), but including ubuntu-advantage-tools in
ubuntu-minimal is not intended to discourage users from upgrading.

> That said, the ubuntu-advantage-tools package is not in fact strictly
> necessary to allow the machine to boot, detect hardware, connect to a
> network, install packages or perform basic diagnostics. With ubuntu-
> minimal 1.400, this is now essentially a mandatory piece of software
> that will, at best, emit an error message.

There are a number of packages pulled in by ubuntu-minimal that could be
removed and still technically fulfill the capabilities listed in the
package's description; they would just make it more inconvenient and awkward
to do so.

ubuntu-advantage-tools certainly falls under the header of 'install
packages' - which is more than can be said of e.g. the tzdata and locales
packages.

As far as emitting error messages: while none of the package services
currently provided by this tool are available on non-LTS releases, it is not
impossible that these would be available in the future depending on
commercial demand.  In the meantime, I think it's a far better user
experience to run the documented command and be told 'this service not
available for this release' than to try to run it and be told 'command not
found', and it was with this in mind that I supported including this package
in non-LTS releases.

> Would changing the relationship from Depends to Recommends meet your
> requirements (automatic setup on every Ubuntu machine) while allowing
> manual removal of the package?

We would have to verify this empirically.  I'm not altogether opposed to it,
but I also don't see what advantage it brings our users.  The
ubuntu-advantage-tools package is:

Installed-Size: 36.9 kB
Download-Size: 11.3 kB

and has zero added dependencies.  And most of the current size is the
archive keys, which if they weren't in ubuntu-advantage-tool would certainly
be in ubuntu-keyring (and indeed, we might still move them there), which is
also part of ubuntu-minimal.

So from a technical point of view, what benefit is there to making the
package removable?

I would note that it would be entirely possible to bundle these scripts with
some other base package; and I think it's really only because this is a
separate package that it attracts notice and invites the suggestion to make
it removable.  Otherwise, having one more command in /usr/bin that goes
unused would not be cause for concern.