Hello Thomas,
On Wed, Sep 07, 2016 at 03:17:31PM +0200, Thomas Goirand wrote:
On 09/07/2016 01:14 PM, Christos Trochalakis wrote:
Package: wnpp
Severity: wishlist
Owner: Christos Trochalakis <yati...@ideopolis.gr>
* Package name : python-confluent-kafka
Version : 0.9.1.2
Upstream Author : Magnus Edenhill <mag...@edenhill.se>
* URL : https://github.com/confluentinc/confluent-kafka-python
* License : Apache-2.0
Programming Lang: Python
Description : Python client to interact with Apache Kafka
python-confluent-kafka wraps the librdkafka C library, providing full
kafka protocol support with great performance and reliability. It's
supported by Confluent, the company behind Apache Kafka.
In which way is this one different from python-kafka? Do we need both in
the Debian archive?
Cheers,
Thomas Goirand (zigo)
On a technical level it is way faster and probably more robust since it
links with the battle-tested librdkafka library.
It is also maintained by Magnus Edenhill, librdkafka's author.
Magnus is now a Confluent employee and Confluent, the company behind
Kafka, is starting to publish clients in multiple languages as part of
Confluent Platform[0]. We can expect that python-confluent-kafka will be
οn par with Kafka feature-wise.
It's true that there is fragmentation in the Python Kafka community,
there is even a GitHub issue about that, opened by pykafka (the third
python client library)[1].
There is no clear winner so I don't see any problem packaging any of
them.
[0] http://docs.confluent.io
[1] https://github.com/Parsely/pykafka/issues/559
