On Fri, 2020-04-03 at 12:45 +0100, Leigh Griffin wrote:
On Thu, Apr 2, 2020 at 3:58 PM Michael Catanzaro
<mcatanzaro(a)gnome.org>
wrote:
> On Wed, Apr 1, 2020 at 8:16 pm, Paul Frields <stickster(a)gmail.com>
> wrote:
> > For a solution to be viable it needs to meet requirements.
>
> Of course, but the problem is that the requirements identified by CPE
> are wildly inconsistent with the actual requirements of the Fedora
> community. The Pagure we have right now seems to be working fine for
> Fedora. All we really need is occasional light maintenance and ensuring
> the infrastructure keeps running. I don't think we'd be having this
> conversation now at all if "dist-git must be open source" was a listed
> requirement, as it should have been from the beginning.
>
> We don't need merge trains or MR approvals or mobile apps (seriously?)
> or private comments or gists or analytics or basically any of the other
> requirements that Neal has lampooned. I understand CentOS and RHEL
> really want merge trains, so maybe that one is a good faith
> requirement, but I honestly don't think most of the rest of them are.
> The list seems to have been concocted by looking at GitLab features
> exclusive to Enterprise and Ultimate editions and then listing as many
> as possible, not by actually listing features that are really actually
> needed to make things work. I know the requirements came from
> stakeholders and not from CPE, but the requirements are so far removed
> from Fedora's actual needs that it has jeopardized the legitimacy of
> the rest of the process. Fedora simply doesn't need any of it. To the
> extent that CentOS or RHEL actually needs a non-OSS feature -- and I
> don't think they *really* do, because those look like nice-to-haves --
> then that just means that their needs are incompatible with Fedora's.
>
> So let's fix the requirements first. Attempting to share requirements
> with CentOS and RHEL has clearly failed.
The CPE team are at the intersection point of CentOS, Fedora and RHEL. We
cannot escape that so any requirements gathered needs to respect that. We
cannot sustain a service of feature for multiple individual stakeholders of
this breadth. This was a driving factor in the exercise and decision.
But if the decision is ultimately to go with a hosted service, there's
no particular reason it needs to be the exact same hosted service for
all three stakeholders, is there? OK, dealing with multiple entirely
different forges could be a drag, but it has been suggested in the
thread that we could ask Gitlab for a hosted Gitlab CE. That doesn't
seem like it'd be difficult.
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net