6:13 a.m.
On Thu, Dec 12, 2019 at 9:40 PM Ben Cotton <bcotton(a)redhat.com> wrote:
https://fedoraproject.org/wiki/Changes/Drop_Optical_Media_Criterion
= Drop Optical Media Release Criterion =
== Summary ==
Proposal to make all Fedora optical media non-blocking. This means
we'd stop blocking on bugs found during the installation of Fedora
from optical media (like CDs and DVDs). This doesn't mean that
installation from optical media would stop working, just that the
Fedora Release wouldn't be blocked on any issues that can pop up in
Fedora installation using this method. Installation from USB devices
will remain blocking.
This has been a long discussion. Let me sum up some answers and
misunderstandings, as a member of the QA team.
The sole reason for the proposal is that testing optical media is a long
and tedious process and we believe that their importance has gone down
massively in the last years. We understand that there are people whose
hardware still requires optical media for installation. There will always
be people like that, even in 10 years. And we don't revel in presenting
them with additional obstacles. But we also have duties, priorities and
limited resources, and must regularly re-assess what we do. In our opinion,
optical media have fallen below the cut-off line. Especially with the
recent deprecation of i386 architecture in terms of kernel/boot support, we
assume that the number of affected hardware and people is just a very small
minority in our user base. That's why we want to drop the release-blocking
requirement and invest the time into testing something that affects more
people.
It's very important to understand that dropping a release criterion for
optical boot doesn't mean Fedora can't be installed from a DVD in the
future. The bugs that affect just bare-metal optical booting (and not
virtual machines or usb booting) are very rare. Adam and Chris have
provided some numbers and references. If such a problem is detected soon
enough, we will definitely do our best to make sure it's resolved by the
official release time. In the worst case scenario, where no one spotted it
in advance and all installation media are affected, there are still avenues
for affected users. An older installation medium can be used and the system
upgraded. And older installation medium can be used and pointed at latest
installation repositories (this is not guaranteed to work, but works in
majority of cases). And we can also spin up unofficial install media
post-release, once the bug is fixed (we've done this in the past). There
are even community members who do these "media refreshes" regularly.
Overall, yes, it might be uncomfortable if you have such hardware, but it's
*not* game over.
It's also important to understand the current state of optical media
release criteria. We've dropped the blocker requirement for most
installation media two years ago. Only Everything netinst and Workstation
Live remained. This proposal suggests that we drop these last two as well.
If you're concerned about Server DVD or KDE Live or something else - there
is no change for you.
Regarding virtual machines, you can rest assured that we'll still block on
VMs booting from ISO files mounted as virtual optical drives. If Adam
thinks this is a bit under-defined at the moment, we'll define it properly
as part of this proposal.
Regarding the topic of how many laptops/workstations/servers sell with an
optical drives nowadays - this is completely irrelevant, let's just not
waste time with this discussion. The important topic is how many of
existing hardware can't boot using other means (that would be mostly USB
for laptops/workstations and network for servers).
Also, this is not something that has been announced suddenly. We've been
discussing this for multiple years, and a year ago it was even proposed by
Matthew Miller (Fedora Project Leader). This gets perhaps more visibility
due to being a Change proposal, as we usually discuss QA matters in test or
test+devel lists. And that's fine, we very much appreciate feedback. I'm
just clarifying this is not some shocking news of the year.
To reply to Justin Flory about Fedora Mindshare Committee - I think this is
more of an engineering decision, really. Of course you'll find users who
will want optical boot 100% supported (and that's true probably about
anything). The question is how much testing we can provide and whether we
want to block the whole release train on such issues, and that's likely
just an engineering prioritization. Of course it's good to have user
feedback.
Regarding Miro Hrončok's proposal "let QA skip testing, but block on it if
somebody else finds the problem" - yes, it's possible. QA is by definition
best effort, even though we've considered a full test matrix coverage
somewhat mandatory lately. We do practice this approach for many criteria
for which we don't even have test cases written. In terms of boot support,
though, I'm not particularly fond of it. Without regular QA testing, this
will have a tendency to get discovered very shortly before Final release,
and then people will be just mad at QA for not detecting it sooner.
If anyone wants to make sure optical media breakage doesn't happen -
please, help us. We announce new candidate composes in test-announce list
regularly. Download the image, burn it, install it, and fill out the
correct field in the installation matrix. Do this from time to time during
pre-branching period, and regularly for every proposed Beta and Final
release candidate. We'll be very grateful. And that is independent on
whether optical media keep being blocking or are no longer blocking. The
test feedback is always helpful. And if the bugs are on our radar, we'll do
our best to have them resolved by the final release day.
Kamil