You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@pulsar.apache.org by Luc Perkins <lu...@streaml.io> on 2018/03/22 18:55:39 UTC

Proposed: Pulsar "instance" terminology change

Hello Pulsar devs,

I've crafted a new PIP for Pulsar, PIP 16
<https://github.com/apache/incubator-pulsar/wiki/PIP-16:-Pulsar-%22instance%22-terminology-change>,
and I'd like to initiate a discussion. The PIP is short and sweet but I'll
provide a quick TL;DR here:

The term *instance* is currently used for a multi-cluster Pulsar
installation. This is an issue for two reasons:

   - It's a somewhat odd terminology choice to begin with and, I think,
   doesn't really comport with ordinary usage of the term instance.
   - The term instance collides with the Pulsar Functions feature. In that
   feature, you can run multiple *instances* of a function (for the sake of
   parallelism). This usage of the term instance is both more in line with
   ordinary usage of the term as well as usage in stream processing domains
   (Heron, Flink, Beam, and others use instance in this way).

So the issue under discussion is (a) whether we should use a different term
for a multi-cluster Pulsar installation, and (b) what that term should be.
I made some suggestions in the PIP, complete with pros and cons for each,
but am 100% open to alternative suggestions.

Luc