kaxil commented on issue #12983:
URL: https://github.com/apache/airflow/issues/12983#issuecomment-762443795


   Again, 100% agree to that @potiuk -- Thanks to you leading that effort we 
are in a much better shape.
   
   But for the issue at hand about the documentation optimisation was what we 
or at least I was talking about. Would definitely love to not break Master for 
that. As I said, I am happy with (3) but we need to be careful with trying to 
optimize where not strictly needed as I mentioned in the previous comment.
   
   All the good work you and everyone who contributed to the CI is very very 
much appreciated by everyone but I think we should now focus on stability and 
the core product.
   
   The number of people who understands the current CI system is in single 
digits and it is very easy to put a blind eye on that one too because you are 
working on it. That is not sustainable in the long run.
   
   Because it is evolving quickly -- it is difficult to understand to keep 
track of the changes -- I would really really stress on the fact that we need 
more stability and more people knowing the current CI and how it works. 
   
   >This is compound change of many things - including the fact that most of us 
can focus on contribution to product, because there are people who constantly 
keep the machinery not only going but also improving and rolling the wheels 
better (even if is more complex).
   
   This is exactly what we should not try to do mainly because currently "the 
people" is just you for major issues (or Ash for some) -- other contributors 
(including me) fix issues here and there but fallback to you.
   
   "I am in favor of introducing these optimizations if we are confident in its 
stability and reliability, and it is also easy to maintain." -- I still 100% 
agree with this statement from Kamil. 
   
   >And any time we found we missed one and it caused faillure we can find it 
and fix it if we find it caused problem (and incrementally get better at 
catching such cases). We will not get it perfect the first time, but it's fine. 
We will catch it an fix it, at the same time vast majority of the build will 
benefit from that.
   
   This is what I don't agree with you about -- there is no need for Master to 
fail for something which is not strictly required.
   
   
   
   
   
   
   


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to