Re: Draft retirement mail for dumpling

[Loic Dachary <loic@dachary.org>]

[-- Attachment #1: Type: text/plain, Size: 3223 bytes --]

Hi Sage,

On 04/05/2015 17:39, Sage Weil wrote:> On Mon, 4 May 2015, Loic Dachary wrote:
>> Hi Sage,
>>
>> Do you think it's worth sending a mail announcing the retirement of 
>> dumpling ?
> 
> Yeah!  
> 
>> It could be something like:
>>
>> Hi,
>>
>> Dumpling, the first major stable release of Ceph[1] published August 2013 retires after more than 18 months of service. People running a Dumpling cluster should upgrade to Firefly and then to Hammer. 
> 
> s/first major stable/major stable/.
> 
> Also, I think we should adjust the timeline page to avoid use of the term 
> LTS (long term supported) and possibly "support" in general as that word 
> means pretty specific things to most people.  Instead, we should say that 
> the developer community backports bug fixes to these releases.  I 
> don't think it is a good idea to make promises about timeframes either.

I think the notion of long term stable releases is something Ceph users rely on at the moment. I'm not including people who mistake "Support" for "Gratis service" in LTS ;) If nothing else the "long-term" in the firelfy & hammer release means something to the reader, to the extent that it probably played an important role in their decision to use Ceph and to chose a specific version. As an informed user of a Free Software I'm well aware there is no contract between myself and the community working on the software. I'm left with guessing and estimating how much I can expect from this community. And my estimation is based on blog posts mentionning "long term", the observed life time of Dumpling, the observed and probable life time of firefly etc. Because of that I think we should make sure the change in terms is not mistaken for a declaration that the upstream will be supported less that it previously was. It certainly is not our intention and what currently happens with the new Stable 
release and backport team is exactly the opposite: we dedicate more efforts to point releases than we previously did.

> Something closer to this?
> 
> 	https://github.com/liewegas/ceph/commit/0415040f1ac6cc22259fc409ed8fc33f86e58c35

That makes sense to me but I think something is missing. The lifecycle of the development release and the stable releases (Emperor, Giant..) is clear. A long term stable release lifetime should extend until two long term support are released. Dumpling ends when Hammer is published, Firefly ends when Jewel is published etc. The rationale being that there are backports to Stable until Stable+1 is published, to fix bugs and possibly backport features. Then there are backports to Stable until Stable+2 is published to fix bugs and focus on whatever can prevent upgrades to Stable+1 so that users have plenty of time to plan for an upgrade and execute it.

I think it is an essential part of the lifecycle of Ceph and a stepstone for all users (individual, non profit, government or companies offering Ceph based services).

With this long term stable release overlap and approximately one long term release per year, 18 months is actually shorter than what will really happen, most of the time.

What do you think ?

> sage
> 

-- 
Loïc Dachary, Artisan Logiciel Libre


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]