[PROPOSAL] ATS v9.0.x

[Leif Hedstrom <zw...@apache.org>]

Hi all,

This was discussed at our last Summit, but here’s what our plan is for v9.0.0:

1) We will not make an effort to merge the ~170 -200 PRs into an v8.1.x release, rather, v8.0.x and possibly v8.1.0 will be trimmed down to bare minimums.

2) We will immediately start focusing on stabilizing master. This goes for everyone, it’s all hands on deck. Essentially, v8.0.x has not been stable since we branched almost a year ago, and we have to avoid this at all cost for v9.0.0. We can’t keep making crappy LTS releases. This work includes looking at the things that we wanted in 8.1.x but could not, as well as fixing v8.0.4 which has not been fixed.

3) We aim to branch v9.0.0 in ~June 2019, with a target of ~October 2019 release. These are not firm dates, and that is a long window, but see 2) above; We have a *lot* of work to do to get v9.0.0 usable.

4) We will start to make releases “goal” oriented, rather than dates. See below for my outline for v9.x. As such, the expectation is that we all work towards the same goals. Of course, the goals are decide by the community, so feedback is required.


We all agreed that this was the way forward. As such, I’m going to Project manage v9.0.x on this Project Page:

	https://github.com/apache/trafficserver/projects/18 <https://github.com/apache/trafficserver/projects/18>


I will make similar pages for v9.1.x and v9.2.x. The tentatively major goals for the upcoming releases are:

	v9.0.0: This brings QUIC into the core, and general server stability!
	v9.1.0: JSON RPC replacing our internal marshaling
	v9.2.0: remap.config to remap.yaml conversion (sidecar), QUIC feature complete
	v9.3.0 / v10.0.0: TBD (suggestions welcome)

These are the general goals for each of these release, we have a few overall goals that we ought to focus on, and I will keep track on those on the Project plan as well. For example, complete migration of configurations to YAML.

Again, this really requires that we all focus on the same general plan! Please speak with your management, and get the buy-in that we need to succeed with this.

Feedback and suggestions always welcome (and expected!)

Cheers,

— Leif and everyone participating at the Beijing Summit

Re: [PROPOSAL] ATS v9.0.x

[Chris Lemmons <al...@gmail.com>]

I really like the idea of goal-oriented releases. I've always felt
that artificial deadlines rarely improve the resultant code. My one
concern is some clarity around how long a given release will be
supported. People who use LTS releases of ATS on account of not having
the resources to keep constantly up-to-date generally need enough
notice to plan ahead for upgrades. (Traffic Control users often check
all those boxes.) So, goal-oriented releases are great, in that if we
hit all our goals quickly, we can get releases into people's hands
fast, but we need to make sure that we maintain the LTS for a
predictable period of time.

> general server stability

Definitely. Stability is by far the most critical feature of ATS.

> remap.config to remap.yaml conversion

I'm glad to see a release or two with a bunch of the new work before
this step. This transition is going to require a fairly significant
update to Traffic Control, so the config format change is definitely a
"big release" for the TC community. In the long term, I'm hoping this
is the last time we have to radically change the config format for
quite a while.

On Mon, Apr 29, 2019 at 3:06 PM Leif Hedstrom <zw...@apache.org> wrote:
>
> Hi all,
>
> This was discussed at our last Summit, but here’s what our plan is for v9.0.0:
>
> 1) We will not make an effort to merge the ~170 -200 PRs into an v8.1.x release, rather, v8.0.x and possibly v8.1.0 will be trimmed down to bare minimums.
>
> 2) We will immediately start focusing on stabilizing master. This goes for everyone, it’s all hands on deck. Essentially, v8.0.x has not been stable since we branched almost a year ago, and we have to avoid this at all cost for v9.0.0. We can’t keep making crappy LTS releases. This work includes looking at the things that we wanted in 8.1.x but could not, as well as fixing v8.0.4 which has not been fixed.
>
> 3) We aim to branch v9.0.0 in ~June 2019, with a target of ~October 2019 release. These are not firm dates, and that is a long window, but see 2) above; We have a *lot* of work to do to get v9.0.0 usable.
>
> 4) We will start to make releases “goal” oriented, rather than dates. See below for my outline for v9.x. As such, the expectation is that we all work towards the same goals. Of course, the goals are decide by the community, so feedback is required.
>
>
> We all agreed that this was the way forward. As such, I’m going to Project manage v9.0.x on this Project Page:
>
>         https://github.com/apache/trafficserver/projects/18 <https://github.com/apache/trafficserver/projects/18>
>
>
> I will make similar pages for v9.1.x and v9.2.x. The tentatively major goals for the upcoming releases are:
>
>         v9.0.0: This brings QUIC into the core, and general server stability!
>         v9.1.0: JSON RPC replacing our internal marshaling
>         v9.2.0: remap.config to remap.yaml conversion (sidecar), QUIC feature complete
>         v9.3.0 / v10.0.0: TBD (suggestions welcome)
>
> These are the general goals for each of these release, we have a few overall goals that we ought to focus on, and I will keep track on those on the Project plan as well. For example, complete migration of configurations to YAML.
>
> Again, this really requires that we all focus on the same general plan! Please speak with your management, and get the buy-in that we need to succeed with this.
>
> Feedback and suggestions always welcome (and expected!)
>
> Cheers,
>
> — Leif and everyone participating at the Beijing Summit
>