Hi Jun,
Thanks for your comment.

These things are never straightforward because of history.

There are broadly speaking the following classes of CLI tools.

1) Console producer/consumer
     Support in-line properties and config files using
     --producer-property and --producer.config, and the
     equivalent for the consumer tools.

     The KIP currently proposes --producer.config and
     --consumer.config are replaced with --command-config.
     It retains --producer-property and --consumer-property.

     Now, --property is already used by the console consumer
     in order to set properties for the formatter. I don't
     think we can move to simply --property for other properties.

     It would be possible to change --producer-property and
     --consumer-property to --command-property.

     There is already --formatter-config, which would align
     with --formatter-property if we chose to use that instead
     of just --property.

2) Verifiable producer/consumer
     When there was just the verifiable producer and consumer,
     using --producer.config and --consumer.config was relatively
     easy to understand because the config was for the producer or
     consumer respectively. However, with the share consumer, the
     admin client is also used, so --consumer.config would have
     been slightly inaccurate and we used --command-config instead.

     This KIP proposes --command-config for all of these.

3) Performance tests
     Support in-line properties and config files using
     --producer.props and --producer.config (and consumer
     equivalents).

     The KIP currently proposes --producer.config and
     --consumer.config are replaced with --command-config,
     and that --producer.props is replaced by --producer-property.

     It would be possible to change --producer-property and
     --consumer-property to --command-property.

4) Everything else
     Support config files only with a variety of flags today.

     The KIP proposes --command-config.

Was this the kind of thing you were thinking? Do you think
--command-property is worth the change? How about
--formatter-property? Or maybe something else?

Thanks,
Andrew
________________________________________
From: Jun Rao <j...@confluent.io.INVALID>
Sent: 13 August 2025 19:28
To: dev@kafka.apache.org <dev@kafka.apache.org>
Subject: Re: [DISCUSS] KIP-1147: Improve consistency of command-line arguments

Hi, Andrew,

Thanks for the KIP.

If we are replacing --consumer.config with --command-config, should we do
the same for --consumer-property for consistency?

Jun

On Wed, Aug 13, 2025 at 9:21 AM Andrew Schofield <
andrew_schofield_j...@outlook.com> wrote:

> Hi Jiunn,
> Thanks for your comment.
>
> Jiunn_00: I was not really familiar with this script, and it's pretty
> user-hostile
> because you cannot see the command-line options unless you choose one
> of the subcommands. This is why it was not in the KIP. When I tried it out,
> it was not apparent that --config existed.  However, I think you're right
> and
> I have added this to the KIP also.
>
> Thanks,
> Andrew
> ________________________________________
> From: 黃竣陽 <s7133...@gmail.com>
> Sent: 13 August 2025 16:16
> To: dev@kafka.apache.org <dev@kafka.apache.org>
> Subject: Re: [DISCUSS] KIP-1147: Improve consistency of command-line
> arguments
>
> Hello Andrew,
>
> Jiunn_00: The kafka-cluster.sh script also has the --config argument.
> Should we
> align this with --command-config for consistency?
>
> Best Regards,
> Jiunn-Yang
> > Andrew Schofield <andrew_schofield_j...@outlook.com> 於 2025年8月13日
> 晚上11:06 寫道:
> >
> > Hi Chia-Ping,
> > It's come to my attention that I completely missed a set of comments
> from you
> > on this KIP. Please accept my apologies.
> >
> >>> chia_00:  Should we introduce `--consumer-property` to
> `kafka-consumer-perf-test.sh` and
> >>> `kafka-share-consumer-perf-test.sh` since another perf tool,
> `kafka-producer-perf-test.sh`, has `--producer-property`?
> >
> > Yes, good idea. Added to the KIP.
> >
> >>> chia_01:  should we align the naming of "how many records"?
> `kafka-producer-perf-test.sh` uses `num-records`,
> >>> while `kafka-consumer-perf-test.sh` and
> `kafka-share-consumer-perf-test.sh` use `messages`.
> >
> > I think this is a sensible alignment. My view is that we generally use
> "record" in Kafka, so I suggest
> > deprecating `messages` in the two consumer tools and replacing with
> `num-records`.
> >
> >>> chia_02:  Have you considering adding `reporting-interval` to
> `kafka-producer-perf-test.sh`?
> >
> > I had not but this is quite a straightforward thing to do. It has a
> reporting interval but just lacks
> > the ability to configure it. I've added it to the KIP.
> >
> >
> > Thanks,
> > Andrew
> > ________________________________________
> > From: Federico Valeri <fedeval...@gmail.com>
> > Sent: 17 July 2025 09:50
> > To: dev@kafka.apache.org <dev@kafka.apache.org>
> > Subject: Re: [DISCUSS] KIP-1147: Improve consistency of command-line
> arguments
> >
> > Hi Andrew, thanks for this useful KIP.
> >
> > This is something that I also though while working on tools migration
> > from Scala to Java, but that wasn't the right time.
> >
> > On Mon, Jul 14, 2025 at 6:18 PM Andrew Schofield
> > <andrew_schofield_j...@outlook.com> wrote:
> >>
> >> Hi Kirk,
> >> Thanks for your email.
> >>
> >> KT01: During the migration period, using both --producer-props and
> --producer-property
> >> is an error. If they use --producer-props, it works and they get a
> deprecation warning.
> >> If they use --producer-property, it works with no warning. I have
> updated the KIP
> >> accordingly.
> >>
> >> I don't really like --command-config either, but I don't think it's
> ambiguous as such. It's
> >> just a bit generic to my way of thinking.
> >>
> >> Thanks,
> >> Andrew
> >> ________________________________________
> >> From: Kirk True <k...@kirktrue.pro>
> >> Sent: 10 July 2025 02:06
> >> To: dev@kafka.apache.org <dev@kafka.apache.org>
> >> Subject: Re: [DISCUSS] KIP-1147: Improve consistency of command-line
> arguments
> >>
> >> Hi Andrew,
> >>
> >> Thanks for the KIP. Yes, the inconsistency drives me crazy :)
> >>
> >> Just one initial question:
> >>
> >> KT01: During the migration period where both command line arguments are
> supported (e.g. --producer-props and --producer-property), which takes
> precedence?
> >>
> >> I'm not ecstatic about the existing naming of "--command-config"
> because it's ambiguous, but that's a battle for another day...
> >>
> >> Thanks,
> >> Kirk
> >>
> >> On Wed, Jul 9, 2025, at 6:53 AM, Andrew Schofield wrote:
> >>> Hi,
> >>> I'd like to start discussion of KIP-1147. This KIP aligns the names of
> the command-line
> >>> arguments across all of the Apache Kafka command-line tools.
> >>>
> >>> KIP:
> https://cwiki.apache.org/confluence/display/KAFKA/KIP-1147%3A+Improve+consistency+of+command-line+arguments
> >>>
> >>> Thanks,
> >>> Andrew
>

Reply via email to