Ubuntu Core 16 timelines

[Jamie Bennett]

Response in-line below.

On 28/04/16 04:19, Mark Shuttleworth wrote:
> Hi folks
> 
> Manik asked me to clarify my comments on priorities in the push to
> release Ubuntu Core 16, the all-snap version of Ubuntu 16.04 LTS.
> 
> This is the "GA" of snappy LTS, so it's a critical step for us. I'm
> absolutely convinced that snaps are an important new primitive for
> operations of Ubuntu-based devices of all sorts, and very focused on
> helping our users have a great experience with them. All the feedback
> we've had on snaps has validated the focus on transactional updates,
> rollbacks, and dependency bundling, which is excellent.
> 
> What we are debating, and my opinion is part of that debate, is what the
> minimal requirements are for us to have a good GA. We are going as fast
> as we can while maintaining quality, the question is just whether we
> take a little longer to have a more complete GA, or whether we put the
> all-snap Ubuntu Core 16 GA out pretty much immediately, at the risk of
> having a more significant upgrade issue during the life of Ubuntu Core 16.

This is a key point. We don't want to put images out there that our
community will build their solutions on expecting it to be solid,
stable, and unchanging. We are aiming for a solid GA in the coming weeks
but we are not at a point where we can hand-on-heart say things are not
going to change and break you.

For our community of application developers we recommend using snapd on
the desktop which gives a great experience and is available today on
Ubuntu 16.04. For people wanting to flash the latest software on their
devices we will be offering *developer* images some time next week. We
guarantee these will change, will break, and do not contain the full
functionality that will be available in the GA but instead are targeted
at our early adopters who accept this caveat.

I will send a follow-up email next week detailing how to download these.

Regards,
Jamie.

> What we know for sure is that Ubuntu Core will get meaningful updates
> over the coming months, in behaviour as well as the standard security
> and reliability fixes. That's because we have a lot of demand for
> capabilities in the snap daemon for policy that isn't there yet, but
> which we have mapped out pretty clearly, and which we will add to snappy
> with updates to Ubuntu Core. For example, enabling better branding of
> webdm for gadget manufacturers, better support for control of updates by
> manufacturers, better support for open source projects that want any
> developer to publish their own snaps of their code, and so on. All of
> these are good things we will get in the next couple of months, through
> updates to Ubuntu Core, and I would very much like to ensure that those
> updates are smooth and hassle free.
> 
> We know pretty clearly that we need to separate the kernel snap, os
> snap, and gadget snap. That's because we know we want to have a free and
> maintained kernel that anybody can use, but we also need gadget
> manufacturers to be able to modify the behaviour of the base device.
> Those modifications go into the gadget snap.
> 
> Gustavo proposed that we publish an image without a gadget snap, or at
> least that was my quick reading of his proposal :) My response was to
> say I would think we need to at least have a minimal gadget snap so that
> we don't have an awkward upgrade. We're having a sprint in ten days time
> where I figure we can hash this out. Anyone with strong insights is
> welcome to join us in Vancouver or share on the list. My comments are in
> the spirit that a two-to-four week delay in publishing official GA
> Ubuntu Core 16 images is worth while if it saves us from having a very
> hard update in three months time.
> 
> I'm in the middle of the OpenStack summit at the moment so am not able
> to participate directly in the debate, but trust the team to make the
> call having considered my opinion. I look forward to hearing all about
> that call in Vancouver :)
> 
> If it isn't yet completely clear, I think snappy is the most exciting
> thing in Ubuntu right now, and that it will make a huge difference to
> everyday usage by developers and device manufacturers, but also people
> who just need to get apps to their VMs in the cloud. Engaging with the
> developer community is our top priority - we have an amazing community
> in Ubuntu, they are very thoughtful and creative and like to find bugs
> and send fixes. My sense is that we can grow the community of people who
> care about snaps fastest by engaging people who are interested in
> building snaps for classic systems (mainly because that's what they and
> their friends have today). All of those community insights help build a
> better snappy, and a better ecosystem of parts for snaps, and a more fun
> community. And that helps us make Ubuntu Core 16 a huge success for
> everybody.
> 
> When Ubuntu Core 16 ships, it will have *exactly* the same libraries and
> binaries and kernel as Ubuntu 16.04 LTS does right now. So encouraging
> people to build snaps on classic Ubuntu 16.04 LTS feeds our snap
> momentum for Ubuntu Core directly.
> 
> I hope this note is reassuring to those who had concerns - I'm not
> suggesting a substantial delay, only that we consider thoughtfully the
> benefits of a slightly more general minimal-viable GA, and if that costs
> us a few weeks now, but saves us from a tricky update later, then we can
> use the time to grow developer participation in snaps on classic. My
> comments are in the spirit of helping us figure out the best plan for a
> huge Ubuntu Core 16, not a diminishing of commitment to the all-snap
> world. Far from it, I seem to have spent all week answer very excited
> questions about Ubuntu Core snaps on top-of-rack switches for large
> scale OpenStack deployments :)
> 
> Mark
> 
>