> Sure, some like big shiny new numbers for new features and new APIs.  But
if we follow the online upgrade-compatibility approach, clusters will be
able to upgrade from any 4.x to 5.1, therefore from the operators PoV the
"6" is not required.
> Aside from my desire to make our semver consistent to just
upgrade-compatibility, I'm in favour of sticking to our general messaging
the past year that Accord will be available in Cassandra 5.  (Introducing a
new major number 6 here IMHO hurts more than helps.)

Thanks for this context and clarification. Based on this I think it makes
sense to follow an "online upgrade-compatibility approach" to define major
versioning and have TCM/Accord on 5.1.

On Mon, Oct 23, 2023 at 11:33 AM Mick Semb Wever <m...@apache.org> wrote:

> Regarding the versioning scheme, if we follow the versioning scheme we
>> have defined "by the book" then TCM/Accord would belong to a 6.0 version,
>> which I have to admit feels a bit weird but it would signal to the user
>> community that a major change is being introduced. I don't feel strongly
>> about this so would be fine with a 5.1 even though it would be a departure
>> from the new versioning scheme we have agreed upon.
>>
>
>
> It can be 5.1 as there's no upgrade-compatibility breakages in TCM/Accord.
>
> Sure, some like big shiny new numbers for new features and new APIs.  But
> if we follow the online upgrade-compatibility approach, clusters will be
> able to upgrade from any 4.x to 5.1, therefore from the operators PoV the
> "6" is not required.
>
> I had a chat with Sam offline about this.  There's a small change in how
> the PropertyFileSnitch works and removing the ability to change a node's
> rack/dc once joined.  It's been suggested to enforce these in 5.0 (with
> simple assertions).
>
> Aside from my desire to make our semver consistent to just
> upgrade-compatibility, I'm in favour of sticking to our general messaging
> the past year that Accord will be available in Cassandra 5.  (Introducing a
> new major number 6 here IMHO hurts more than helps.)
>
>

Reply via email to