potiuk commented on issue #10753:
URL: https://github.com/apache/airflow/issues/10753#issuecomment-687708299


   I believe the XKCD image (which I also retweeted) is very apt and it 
describes the situation in both cases:
   
   a) either you depend on a binary that someone builds and maintains
   b) or on the sources that someone maintains
   
   > That we should not make it mandatory that all the sources needs to be 
under our umbrella if sources are clear enough and licensing is not the issue. 
My main concern is any code in our repo if we don't maintain (keep up to date) 
it will become technical debt.
   
   I think I already agreed to it? Didn't I. I got convinced that forking and 
releasing the whole repo is not necessary. I don't think this is what we argue 
about anymore.
   
   As long as we give the users a clear way how to build the image on their own 
using "official binaries" and "sources" that are likely to be still there if 
they look for them.
   
   > If there is a lack of docs on how to rebuild it, let's help the OSS 
community, even if it is a small project. I definitely believe in the ASF motto 
of community before code, we could help the community (however tiny of that 
exporter) by letting them know that this is the issue and if they can address 
the rebuild instructions by just adding some clear docs on it.
   
   We have no control about this. I am happy to open a PR to add more clear 
docs there, still we have to tell our users where they can find the code to 
build the image. But I want to not to be dependent on someone else's decision 
on that. Just an example - I usually do this. Not sure how often you do that 
but - I pretty much always do that when I see an opportunity, I've done this 
very thing twice recently:
   
   * https://github.com/hashicorp/setup-terraform/pull/38 - > modification to 
Github Action of  Hashicorp's Terraform to be able to add < version constraint. 
We needed it as Terraform released 1.13 with breaking changes. I've added all 
the tests and docs and still wait for a single word of feedback after nagging 
them several times
   
   * https://github.com/n1hility/cancel-previous-runs/issues/7 - > issue which 
I opened with the original author of the  "cancel-previous-run" Github Action 
which was created before 'workflow_run' got available. I propose the author 
that we either deprecate his GA or they can merge my changes and take it over. 
Still no answer after 17 days and nagging the author.
   
   Those two are two different types of communities - one big and established, 
one small and 'niche'. But I am still waiting (not for acceptance, but for 
feedback at all to my kind proposals). So while I am perfectly OK with asking 
the author to improve the process, I have no high hopes it will happen. And 
(similarly as in the case of those two PRs above) I would like to have a way I 
can proceed on my own. I can promise you that once we agree on how WE are going 
to proceed, I will open a PR or an issue for the pgbouncer-exporter with my 
proposal. But I want to proceed on our own regardless of the answer.
   
   > Regarding the technicalities of this exporter I will leave it upto you, 
Ash and Daniel as long as we are clear that this is not a requirement if the 
sources are clear enough and licenses are ok for future such items.
   
   What I deeply care about is about our users. I want our users to get clear 
instructions from us on how they can rebuild the image (as mandated by the 
release policy they should be able to do it) - using official binaries and 
sources that are both legally available to them.  
   
   Please correct me if I am wrong but I believe they did not have this 
information before. They do have it with the #10759.  
   
   Basically we give our user this choice:
   
    a) each user spends a couple of hours or days to find out where the sources 
come from and how to rebuild the image
    b) they got a script and Dockerfile that does it (very similar to what we 
do with the INSTALL  instructions for airflow itself)
   
   If you do not agree with the script, could you please propose PR which 
solves it differently in the way that - as a user - I do not have to make an 
investigation on how to build the images. I am really looking for a 
constructive way to solve this and so far I have only my PRs and code against 
("let's simply point to the images and let the user figure out how to rebuild 
them if they need to") which I think is unacceptable.
   
   Can you propose another solution @ashb @dimberman @kaxil ?
   
   
   


----------------------------------------------------------------
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