potiuk edited a comment on pull request #19361:
URL: https://github.com/apache/airflow/pull/19361#issuecomment-957490564


   I will be able to take a closer look later this week. I have not yet looked 
at it, but it seems that this feature is something that we need a very good 
look at adn make sure we understand the consequences of it especially when we 
consider future ideas of being able to make Airflow more 'task-oriented" than 
"Dag-oriented" which we know is something we want to work on.
   
   Again - I have not looked at it so it is hard to comment yet and I do not 
know the scope and consequences of that, But I generally see three options we 
could treat this kind of change (since it was not discussed in an AIP/devlist 
before:
   
   a) if it will fit into the long term idea and is an incremental change 
without foreesen big impact on target architecture we "envision" (even if it is 
not specified yet) we could iterate on it and merge it as "tactical" solution 
for task scheduling
   
   b) if it will be risky - in terms of blocking some ideas and directins - we 
should discuss those and see if the ideas are viable/whether we can turn it 
into a).  If we cannot and the risk is too big (and we foresee to work on more 
"task-based" scheduling soon then this PR might be good one to discuss stuff 
and eventull go to solution c)
   
   c) we could start discussing more "fully featured" solution for task-based 
scheduling and start discussin in the devlist and start drafting AIP - in which 
case this PR might serve as a good "starting point" for discussion - either as 
an example of things we want to avoid and implement differently or showing 
where we still need to get missing puzzles to fit in order to know that the 
direction we take is good.
   
   I have no idea whatsoever which of the options above it can fit in - it can 
be either, but I think it's good to agree and let @malthe know soon from the 
committers who would be interested, which direction is most likely here (I am 
at my hometown today and coming back tomorrow - so i will take a look some time 
later this week) - but also I'd love to hear what others who thought about the 
"task-based airflow" to state their opinions here.


-- 
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.

To unsubscribe, e-mail: [email protected]

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


Reply via email to