Kafka is first-rate, rock-star technology, just as is NiFi.

It would be nice to find something from Kafka elaborating on how this regular and accelerated release cadence is working out for them, how much more work it's been, what problems they've experienced, etc.

I show their releases over the last couple of years as below[1]. The cadence appears to be settling into the the 4-month cycle proposed. It's possible to discern a maintenance schedule. It doesn't exactly match NiFi's 0.x and 1.x efforts (which were simultaneous for some time too), but it's clear they've faced similar complexity (maybe a little more though for a shorter time). And, of course, there's no meaningful way to compare the effort going into and features implemented in Kafka by comparison with NiFi.

2019
2.3.1    24 October
2.3.0    25 June
2.2.1     1 June
2.2.0    22 March
2.1.1    15 February

2018
2.1.0    20 November
2.0.1     9 November
2.0.0    30 July
1.1.1    19 July
1.0.2     8 July
0.11.0.3  2 July
0.10.2.2  2 July
1.1.0    28 March
1.0.1     5 March

2017
1.0.0     1 November
0.11.0.1 13 September
0.11.0.0 28 June
.
.
.

[1] https://kafka.apache.org/downloads

On 11/5/19 8:02 AM, Pierre Villard wrote:
Hi NiFi dev community,

We just released NiFi 1.10 and that's an amazing release with a LOT of
great new features. Congrats to everyone!

I wanted to take this opportunity to bring a discussion around how often
we're doing releases.

We released 1.10.0 yesterday and we released 1.9.0 in February, that's
around 8 months between the two releases. And if we take 1.9.2, released
early April, that's about 7 months.

I acknowledge that doing releases is really up to the committers and anyone
can take the lead to perform this process, however, we often have people
asking (on the mailing lists or somewhere else) about when will the next
release be. I'm wondering if it would make sense to think about something a
bit more "planned" by doing time based releases.

The Apache Kafka community wrote a nice summary of the pros/cons about such
an approach [1] and it definitely adds more work to the committers with
more frequent releases. I do, however, think that it'd ease the adoption of
NiFi, its deployment and the dynamism in PR/code review.

I'm just throwing the idea here and I'm genuinely curious about what you
think about this approach.

[1]
https://cwiki.apache.org/confluence/display/KAFKA/Time+Based+Release+Plan

Thanks,
Pierre


Reply via email to