On Wed, 2015-06-03 at 08:43 -0400, Stephen Gallagher wrote:
In addition to the PRD refresh that we are all looking at (*ahem*),
we
also need to figure out what are the major efforts we're going to
engage in during the Fedora 23 cycle. Alpha Freeze for Fedora 23 will
be on July 28th[1], and that means we need to have at least a
prototype
of what we want in Fedora 23 ready by that point.
I'll list some of the things that have been discussed on the mailing
list and in meetings recently and we can work from there. (I'll also
include my recommendations inline).
==============================
== File-sharing server role ==
==============================
=== Benefits ===
* Simple, easy-to-use mechanism for producing a file-sharing host
using the CIFS and/or NFS protocols.
* Clear gap in our offering today.
=== Requirements ===
* Development of an API to manage the configuration.
* Creation of a server role definition to provide the services.
* Graphical interface for manipulating it (likely Cockpit, unless
another web-based tool can be adapted without sacrificing usability).
=== Issues/Risks ===
* Cockpit developers have indicated they cannot spare the resources
to
implement the GUI this cycle.
=== Recommendations ===
Ask for volunteers to work on the management API for Fedora 23 and
postpone cockpit and rolekit work until Fedora 24.
============================================
== Cockpit GUI for Domain Controller Role ==
============================================
=== Benefits ===
* Easier setup of initial domain controller.
* Demonstration of our commitment to converging capabilities into
Cockpit
=== Requirements ===
* UI Design of role creation (started; thanks Andreas Nilsson!)
* Resources familiar with JavaScript and D-BUS (acquired: we have a
GSoC student working on this)
=== Issues/Risks ===
* Work is being done by a student as part of Google Summer of Code.
Student projects always carry some risk. GSoC schedule is a good fit
with the Fedora 23 schedule (recommended pencils-down date is before
Fedora 23 Beta)
=== Recommendations ===
Let's definitely move ahead with this one. I'm sure our student
(Turner
England) will be motivated and excited to see his work featured
prominently as a deliverable. I will also be personally involved as
one
of his two mentors (the other being Peter Volpe of the Cockpit team).
================================
== Containerized Server Roles ==
================================
=== Benefits ===
* Server roles offered by containers will be unaffected by
underlying
OS upgrades.
* Containers written under the Nulecule specification will be
possible
to migrate to the Fedora Atomic Platform (and possibly also OpenShift
v3) later.
=== Requirements ===
* At least one proof-of-concept role (I am proposing memcached for
this, as it's very simple).
* Ideally convert the Database Server role to a container format.
Must
handle upgrades (or backup/restore) from F22 DB Role.
* Docker package must be built for at least all primary arches (and
ideally the secondary arches as well).
=== Issues/Risks ===
* More difficult to determine when upgrades are necessary and to
apply
them.
* Docker Hub only supports x86_64, so we will have to either provide
a
hub of our own or build docker images in rolekit from Dockerfiles.
=== Recommendations ===
I will be working on at least the PoC role (memcached) at the
direction
of my management at Red Hat, so we should probably treat this as a
given and go ahead with filing at least one Change Proposal.
I recommend for simplicity (and not adding to rel-eng's workload)
that
we build the images in rolekit from Dockerfiles based on the latest
available RPMs in the standard Fedora package collection.
I also recommend that we defer attempting to completely solve the
docker image upgrade problem to Fedora 24 (there's some work going on
in the Atomic space that may be more mature by that point). I think
for
a first attempt, it will be acceptable to just build the ability to
perform upgrades on demand and instead implement the notification of
upgrades at a later date.
==========================
== FreeIPA Replica Role ==
==========================
=== Benefits ===
* Ability to set up a resilient domain
* Fewer manual steps after the initial domain creation
=== Requirements ===
* Mechanism to promote a domain client to a replica
* Resources: This needs the heavy involvement of a FreeIPA developer
(simo?)
=== Issues/Risks ===
* May require plumbing changes that cannot be completed in time.
=== Recommendations ===
Feedback needed from FreeIPA team before I can make any
recommendations.
[1]
https://fedoraproject.org/wiki/Releases/23/Schedule
Some items I forgot yesterday:
==============================
== Stable API Documentation ==
==============================
=== Benefits ===
Having a Fedora-provided repository of the specific set of API (and
ABI) that we guarantee stability or backwards-compatibility will go a
long way towards addressing the concerns around our lack of an LTS
release. Additionally, it will make life easier for our users to have a
single source to look for documentation, rather than the current
situation of having to search out each upstream for documentation.
=== Issues/Risks ===
* Requires a large time commitment from someone on the Fedora Docs
team to collate the documentation and post it to the Fedora
Documentation site.
* Requires developer effort to locate and identify the stable APIs
* Almost certainly cannot all be done in a single release
=== Recommendations ===
* Locate someone from Fedora Docs to do the collation
* Start with a set of known-stable APIs (such as glibc and systemd)
and publish those for Fedora 23.
=========================
== API break detection ==
=========================
=== Benefits ===
Provide a taskotron process that will identify API and ABI breaks for
common languages when updates are submitted to Bodhi. If such are
detected, we should disable autopush-by-karma. This will allow us to be
able to better avoid incompatible updates in stable releases of Fedora.
=== Issues/Risks ===
* Tooling needs to be implemented. (Some help may be available from
libabigail[1])
* Someone needs to write a taskotron process that will run when
updates are created
* Available tools for this are currently limited to C/C++ ELF
libraries
=== Recommendations ===
Search out someone to do this work. It is high value for comparatively
little work (since much of the hard work has been solved by
libabigail). Volunteers highly requested.
It would be worthwhile to start with C/C++ and see how things progress
from there.
[1]
https://sourceware.org/libabigail/