An SPIP is technically a Jira ticket with a special label 
<https://github.com/apache/spark-website/blob/d3ce43b45104eab170b0c383ed5b74f365ec6c28/improvement-proposals.md?plain=1#L18-L22>.
 The SPIP proposal and associated discussion are sometimes on the ticket 
itself, but more often they are in a linked Google Doc.

I think the main reason SPIP authors like using Google Docs is because it 
enables line-level comments and threaded discussions, which Jira does not 
support. But we already have a mechanism for this: the pull request.

Python enhancement proposals (PEPs) are hashed out on PRs and ultimately stored 
in git <https://github.com/python/peps/tree/main/peps>. (For consensus, they 
use a Discourse forum. This would be analogous to our dev list. I am not 
proposing changing this aspect of the SPIP process.)

Why don’t we do something similar for SPIPs? I don’t see why we should accept 
the mix of Jira ticket descriptions, Jira comments, and Google Doc comments, 
when an SPIP can simply be a Markdown document submitted via PR. When an SPIP 
is accepted via the usual vote process, the associated PR is merged.

All accepted SPIPs would live in one place as a folder of Markdown documents. I 
suggest a new `spips/` folder in `apache/spark-website`. They would still be 
associated with Jira tickets. They don’t have to be rendered on the website. 
And we don’t have to migrate prior SPIPs.

What do you think?

Nick

Reply via email to