Thank Colin.

> One last note: if we do this, we should pass the UUID of the topic as
well as the name.

Sure, adding it to the KIP.

> On the minus side, the stated use-case (preventing deletion of
__consumer_offsets) seems easy to solve via ACLs. The CreateTopics case is
different... it's not as easy to solve via ACLs because people wanted to
enforcce specific topic names or conventions, beyond what ACLs could
provide.
> So it would be good to understand a bit more about why ACLs are not a
better solution than deletion policies.

Thanks for highlight this. I elaborated on the motivation a bit further.
I agree ACLs are a main tool to protect against unwanted topic deletion;
but even when proper users are authorized, it may be a human error to
request a topic deletion. So, in this case, policies act as a complement to
ACLs when topic deletion wants to be blocked.

Looking forward to your feedback.

Many thanks,
Jorge.

On Fri, 2 Jun 2023 at 22:24, Colin McCabe <cmcc...@apache.org> wrote:

> Hi Jorge,
>
> On the plus side, the change is small and pretty easy to support.
>
> On the minus side, the stated use-case (preventing deletion of
> __consumer_offsets) seems easy to solve via ACLs. The CreateTopics case is
> different... it's not as easy to solve via ACLs because people wanted to
> enforcce specific topic names or conventions, beyond what ACLs could
> provide.
>
> So it would be good to understand a bit more about why ACLs are not a
> better solution than deletion policies.
>
> One last note: if we do this, we should pass the UUID of the topic as well
> as the name.
>
> best,
> Colin
>
>
> On Mon, May 22, 2023, at 09:18, Jorge Esteban Quilcate Otoya wrote:
> > Hi everyone,
> >
> > I'd like to start a discussion for KIP-934 <
> >
> https://cwiki.apache.org/confluence/display/KAFKA/KIP-934%3A+Add+DeleteTopicPolicy
> >
> > which proposes adding a new policy for when deleting topics.
> >
> > There have been related KIPs in the past that haven't been accepted and
> > seem retired/abandoned as outlined in the motivation.
> > The scope of this KIP intends to be more specific to get most of the
> > benefits from previous discussions; and if previous KIPs are resurrected,
> > should still be possible to do it if this one is adopted.
> >
> > Looking forward to your feedback!
> >
> > Thanks,
> > Jorge.
>

Reply via email to