[MXS-3903] Investigate Postgres wire protocol Created: 2021-12-07  Updated: 2022-08-15  Resolved: 2022-03-30

Status: Closed
Project: MariaDB MaxScale
Component/s: Core
Affects Version/s: N/A
Fix Version/s: 22.08.0

Type: Task Priority: Major
Reporter: Johan Wikman Assignee: Niclas Antti
Resolution: Done Votes: 0
Labels: None

Sprint: MXS-SPRINT-146, MXS-SPRINT-147, MXS-SPRINT-148, MXS-SPRINT-149

 Description   

Take a look at the Postgres wire protocol and provide a rough estimate of what kind of effort would be needed to support it.



 Comments   
Comment by Niclas Antti [ 2022-08-15 ]

from #postgressql-friendly

Hi everybody. I looked into PG a few months back. Reading the messages above, I agree with the notion that if there were to be PG compatibility in MariaDB it should be limited to the subset seen as essential to PG users. The scope of my research was very limited and I quickly realized that most of the questions and issues to solve are on the MariaDB side, or maybe even more so on the business and marketing side. We can re-prioritize PG in MaxScale if there is a need for that, but I think this forum is good as the "What" needs to be researched before the "How".
You are right about the wire protocol, it is not the hard part. A straight protocol converter in MaxScale could be implemented in relatively short order. Simple PG extensions could be translated in that module.
In case more of the PG dialect and extensions had to be supported in MaxScale, we might have to use their parser (rather than enhance our own), which obviously would take a bit longer.
Overall I am somewhat against implementing PG features in MariaDB and MaxScale, and more for migration tools, be it that we add a module for migration in MaxScale as long as it can be defined so that it does not cause the rest of the code base to be touched to any greater extent.

Generated at Thu Feb 08 04:24:47 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.