On Tue, Mar 8, 2016 at 8:52 AM, Stephen Gallagher <sgallagh(a)redhat.com> wrote:
On 03/07/2016 09:34 PM, Josh Boyer wrote:
> On Mon, Mar 7, 2016 at 9:03 PM, Stephen Gallagher <sgallagh(a)redhat.com>
>>> On Mar 7, 2016, at 8:37 PM, Matthew Miller <mattdm(a)fedoraproject.org>
>>> On Tue, Mar 08, 2016 at 01:30:39AM +0000, Peter Robinson wrote:
>>>>> Ideally, we'd have roles to cover just about _anything_ anyone
>>>>> wanted, and so the process would be start with Fedora Server
>>>>> _instead_ of Minimal, and then install your role, instead of
>>>>> packages, but maybe that's too far out.
>>>> Also at the moment server is all that is built for Secondary arches and
>>>> so with a DVD installer it is the only way they can get a minimal.
>>> There's been some talk recently of having a re-chartered Base WG
>>> a functional minimal install (as well as the container base image). That
>>> seems like it might serve some of the Secondary arches better (although
>>> for others, actual Server seems appropriate too). And, I guess that's
>>> still separate from whether Server should _also_ present a minimal
>>> environment of its own — that's a question of audience and use cases, I
>> I'm of the opinion that Server Edition should realistically be as minimal
>> as we are willing to support. That's not the same thing as the
>> build-your-own-OS case of the minimal environment today.
> It is unclear whether you are saying the "Base WG maintain a functional
> minimal install" is a bad idea, or if you are simply saying the current
> Server image is as small as you want to go for Server. The we is also
> unclear. We as in Fedora or we as in Server?
Neither of those was what I was trying to say, actually. What I was saying is
that my view of Server Edition is that it should be designed to be as minimal as
possible without compromising a basic level of functionality. The end goal is
for the default Server Edition installation to meet the needs of at least a
majority of users running infrastructure workloads. Focusing on an absolutely
minimal footprint has been explicitly a non-goal of the Server SIG, because we
want to deliver an actual product, not an erector set.
Now, this is a delicate balance to reach and it's possible (even
likely) that we
haven't struck it yet. What we need are the actual use-cases of those people
that are still insisting on the "minimal" environment. We're looking for
answers too, not just cargo-cult: "I want it to be smaller". I want to know
actual reasons why smaller is *better*. (I understand some of them, but most of
the ones I know are generic and fall broadly into "fewer packages means fewer
updates" and "fewer running services means a reduced potential attack
Beyond those, I would want to hear real-world use-cases for why we should offer
a server that can ship without specific functionality. The most common ones I
hear is people wanting to ship without Cockpit or firewalld because they have
their own management solutions. But at the same time, we've very specifically
defined from the beginning that these two things constitute part of the
guaranteed API of Server Edition in order to make it broadly useful. (I could
see us offering a "headless" version of Server where we didn't ship the
cockpit-ws package, thus making it manageable only remotely from another server,
Right, all of that makes sense.
> Sorry, I'm not trying to be difficult. I simply don't
> what your point of view here is.
>> If Server Edition isn't small enough for some use cases, then we should
>> identify those specifically and decide if it's worth adjusting the contents
>> of the Server Edition to handle that case or if it belongs as part of
>> another (perhaps new) Edition.
> The often spoke of Base WG minimal image is thought of as the bare bones
> minimal image that everything else is built on top of. This ties directly
> into modularity in that the "image/container" is a functional check of the
> minimal modules. Hopefully it becomes more clear after the meeting
Well, I don't think we ever decided if Base was supposed to produce an
installable artifact. (I think it never did, since the Minimal image is the most
likely candidate and that hasn't seen any curation for the last couple years
that I can see).
Right, it was undefined. Now we're trying to move forward and, to put
it bluntly, revitalize Base both from a WG and from an "what do they
I agree that we need to hear more at the meeting tomorrow, after
have a better idea of what Minimal should be and whether the Base WG should be
I don't even really like calling it Minimal. In my head, if the Base
WG produces an artifact (container, iso, repository, etc), it isn't a
Capital E Edition. It's likely more "here's a thing we did that can
be used to build on top of to create something more useful." That
artifact is a byproduct of the actual things Base would care about,
and only used for testing purposes. Those things may be dependency
minimization and/or creating a set of packages that constitute the
fundamental building block everything else needs. If the artifact is
useful on its own for some other group or purpose, that is a bonus.
Even that is a fuzzy definition, so let's discuss during the meeting
if it comes up.