+ 1 (binding)

I also like Jarek's proposal. airflow CLI is way too powerful to expose to
the average user and ctl approach is best.
However we shouldn't restrict admins in a way that they need install ctl to
access api supported commands.

A good example is local development with airflow standalone,
there is no need to have a seperate airflowctl configured for your scratch
dev environment. (+2 for Jareks suggestion :D)


On Tue, Jan 27, 2026 at 3:34 PM Jarek Potiuk <[email protected]> wrote:

> +1 (binding).
>
> But I have one possible improvement proposal (this might be done as a later
> amendment if we find good support).
>
> I am a little concerned about the migration "stage" when some "airflow"
> commands will raise deprecations, some will be missing, some will be still
> there (but are going to be removed) and some will be there without plans to
> be removed. This might get people really confused on what they should use
> to run some commands.
>
> I think we would still get exactly same benefits for maintenance, if we do
> NOT deprecate nor remove `airflow` commands from "airflow" CLI at all  -
> but simply **silently** redirect them to airflowctl under the hood
> (airflowctl will have to become a dependency of `airflow-core`). The
> authentication could then be done "on the flight" - without even the user
> storing any user:password locally - `airflow` could dynamically generate a
> token for the remote call (it already has all the necessary config to do
> so).
>
> This way we would not have to have dedicated users created and
> user/passwords stored locally for "airflow" commands to work - essentially
> keeping **exactly** the same behaviour of "airflow cli" when you run them
> on the "airflow-core" machine. Documentation could be generated from that
> single source of truth in airflow-ctl (but otherwise it would remain the
> same for all the remote "airflow" commands). It would then be just a
> technical change from directly calling python API for the CLI commands into
> running airflow-ctl local HTTPS API calls but for users things would remain
> the same.
>
> I think it has **exactly** the same benefits (no redundancy in command
> maintenance). And it might ease the learning process (basically NO learning
> needed at all - everything will work exactly the same as today) - and allow
> truly incremental migration. And there is no need whatsoever in this case
> to remove the "airflow" commands. They will **just work** as if
> `airflowclt` was called.
>
> We are **anyhow** going to have "airflow" cli for the local admin tasks, so
> the CLI needs to be there, I see generally no drawback in leaving the
> "empty shell" airflow commands to call airflow-ctl automagically. I think
> it would be super convenient for the person who does admin tasks on
> "airflow-core" to run the same "airflow" CLI for all kinds of commands
> rather than remembering which ones are in "airflow-ctl" and which ones are
> in "airflow" CLI.
>
> Again - not a blocker and this is still my +1 for the original proposal -
> but maybe a good opportunity to get the voting people (and Buğra of course)
> to say what they think about this kind of "amendment".
>
> J.
>
>
> On Tue, Jan 27, 2026 at 9:07 PM Jens Scheffler <[email protected]>
> wrote:
>
> > +1 (binding)
> >
> > On 27.01.26 20:16, Buğra Öztürk wrote:
> > > Hey all,
> > >
> > > I would like to call a vote on this AIP
> > >
> >
> https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=382175838
> > >
> > > The previous discussion thread was
> > > https://lists.apache.org/thread/rgbbl8dt79nt455mh62m34dcj2b9s0lb
> > >
> > > Based on the discussion on AIP-81
> > > <
> >
> https://cwiki.apache.org/confluence/display/AIRFLOW/AIP-81+Enhanced+Security+in+CLI+via+Integration+of+API
> > >,
> > > this has already been agreed upon as a plan.
> > > We have a more concrete plan to reduce the impact on users as much as
> > > possible on AIP-94.
> > > It is not a child AIP. It contains two different projects, which
> include
> > a
> > > tooling and a plan for migration from Airflow CLI (Remote Commands, see
> > > AIPs) to Airflow CTL (airflowctl).
> > >
> > > Thank you to everyone who commented on the AIP and responded to the
> > > discussion thread!
> > >
> > > The vote will run for 5 days and last till next Sunday, the 1st of
> > February
> > > 2026 21:00 CET.
> > >
> >
> https://www.timeanddate.com/countdown/election?iso=20260201T21&p0=16&msg=%5BVOTE%5D+AIP-94&font=cursive
> > >
> > > Everyone is encouraged to vote, although only PMC members' and
> > Committers'
> > > votes are considered binding.
> > >
> > > Please vote accordingly
> > >
> > > [ ] +1 Approve
> > > [ ] +0 no opinion
> > > [ ] -1 disapprove with the reason
> > >
> > > Please consider this is my +1 binding :)
> > >
> > > Regards,
> > > Bugra Ozturk
> > >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: [email protected]
> > For additional commands, e-mail: [email protected]
> >
> >
>

Reply via email to