On May 3, 2018, at 12:42 AM, Yaniv Kaul <ykaul(a)redhat.com>
wrote:
Patches are welcome to improve the way oVirt uses Postgresql,
supports various versions, etc.
Can you give examples for some of the things you'd do differently?
A little pre-ramble - I was trying not to be offensive in talking about this, and hope I
didn't bother anyone. For the record, if I were supreme dictator of the project, I
might well make the same choices. Attention is limited, DB-flexibility is nowhere near a
top-line feature, DB compatibility issues can be complex and subtle, and QA is a limited
resource. I don't know that those are the concerns responsible for the current stance,
but can totally see good reasons as to why things are they way they are.
Anyway, I've been thinking about is an installer mode that treats the DB as Someone
Else's Problem - it doesn't try to install, configure or monitor it, instead
leaving all config and responsibility for selecting something that works to the
administrator. The assumption is that crazy people like me will figure out if things
won't work against a given version, and over time the list will be capable of assuming
some of that QA responsibility. That leaves the normal path for the bulk of users, and
those who want to assume the risk can point at their own clusters where the closest
running version is almost always going to be a point-release or three away from whatever
Ovirt tests against and configuration is quite different.
What I have not done is written any code. I'd like to, but I'm probably several
months away from having time.
-j