PMC Bandwidth Updates

[Victor Barua]

At the last sync [1] we talked a bit about limitations in PMC bandwidth for reviews. This primarily stems from the fact that at this point in time only 2 PMCs are actively remunerated for work on this project.

However, the work must go on, and with the group we brainstormed about how we might tackle this. A number of good ideas were suggested like:

• Adding more PMCs.
• More fine-grained voting policies around what does and does not need PMC review.
• Having a PMC to Committer vote equivalence.
• Leaning more heavily on our pre-1.0 status and being more comfortable with experimentation and breaking changes.

This last one, especially, resonated with me. Many of the things that we're trying to capture in Substrait don't really have right or wrong answers, and the best way to determine if something works sometimes is to just try it. As Weston pointed out, "Analysis Paralysis" is a real issue. Historically, we've tried to move slowly and carefully, but there is such a thing as over-analysis. Instead of being afraid to make mistakes, let's make it easier to experiment and roll them back. In this spirit, I've opened up a PR to document a breaking change policy.

I do want to say that I appreciate the work that all of our PMCs have done, and continue to do, with whatever capacity they have available. Thanks for getting us to were we are today, and here's to many more years of Substrait.

[1] Discussion Recording

[Jacques Nadeau]

I unfortunately missed this weeks sync to a family emergency so thanks for bringing this to the mailing list. 

Another option is figuring out ways to get our less active once more engaged…

--
You received this message because you are subscribed to the Google Groups "substrait" group.
To unsubscribe from this group and stop receiving emails from it, send an email to substrait+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/substrait/7d7b13dd-4847-4f0a-9848-dfc0fe8553ffn%40googlegroups.com.

[Ben Bellick]

I agree that getting more community engagement would be great! However, the biggest bottleneck for us is in getting PRs merged in. Adding more contributors would help generally with community engagement, but we need to figure out how to stop the backlog of open PRs from growing. 

Any of the things Victor suggested above would help a lot :) 

To view this discussion visit https://groups.google.com/d/msgid/substrait/CAJ9XdSpW2Es%3Deh9x-6QgTHBrNZQi%2B-zRpsC2J8sam41Nkr-w_A%40mail.gmail.com.

[Jacques Nadeau]

My thoughts on the proposed items thus far:
> Adding more PMCs.

+1

> More fine-grained voting policies around what does and does not need PMC review.

Not convinced. What concrete changes do you think would be good? Are there specific things we see as specifically restrictive/not right?

> Having a PMC to Committer vote equivalence.

-1. Not a fan.

> Leaning more heavily on our pre-1.0 status and being more comfortable with experimentation and breaking changes.

Same challenge here, not sure what the concrete change would be.

> Re-energize existing PMCs

+1. I'm going to work to help out more.

I don't think that the community has suffered from analysis paralysis. I think what we're doing is just hard. When we started I remember talking to the folks that worked on the orca optimizer which (The closest prior art to what we're trying to do.)  The "universal optimizer" quickly became an optimizer "for these two projects only." It's just so easy to focus on  "making things work for me". It's much harder to always try to make things work for many systems. Tickets often suffer because the proposer hasn't fully considered the big picture (often just due to a lack of familiarity with the problem space). This puts a much larger burden on reviewers. I find reviewing patches on substrait often harder than other things.)

I think we have suffered a lot from drive-by attention. I feel like I'm a big offender on that front. That may look like analysis paralysis but I think it's a different problem.

To view this discussion visit https://groups.google.com/d/msgid/substrait/CAFKh94CpOpU%3DK33K05Lkoe5WCarJHsqsTJgBxg19QM9OZtruUA%40mail.gmail.com.

[Victor Barua]

> > More fine-grained voting policies around what does and does not need PMC review.

> Not convinced. What concrete changes do you think would be good? Are there specific things we see as specifically restrictive/not right?

Maybe fine-grained voting policies was the wrong phrasing. I can think of small improvements to the governance docs to improve clarity like:

* Clarify that deprecations require PMC votes.

* Clarify that CI / Build updates in the core repo just require committer approval.

I, personally, didn't have anything particularly big policy changes in mind, but small clarification would help folks excercise their judgement.

>> Leaning more heavily on our pre-1.0 status and being more comfortable with experimentation and breaking changes.

> Same challenge here, not sure what the concrete change would be.

> I don't think that the community has suffered from analysis paralysis.

I can say that I personally have, because I worry about missing edge cases in systems I'm not familiar with, or weird interactions with existing features. At a certain point I/we have to trust that we've done enough of the leg work and be okay with revisiting things if we realized we were wrong.

> I think what we're doing is just hard ... It's just so easy to focus on  "making things work for me". It's much harder to always try to make things work for many systems. Tickets often suffer because the proposer hasn't fully considered the big picture (often just due to a lack of familiarity with the problem space). This puts a much larger burden on reviewers.  I find reviewing patches on substrait often harder than other.

I also agree with this though. I've felt like the onus has been on reviewers to check the big picture, but maybe we/I need to be more comfortable pushing back on contributors to do some of the leg work.

All that said, I think we're trending in the right direction at this point, especially with potentially new PMCs incoming.