FYI: I had emailed Antoine last week and he is away this week. He should be able to reply next week.
On Tue, Jul 14, 2026 at 1:00 PM Ryan Blue <[email protected]> wrote: > Sorry, I sent this on Friday but forgot to change the sender to my ASF > account and my gmail gets blocked for some reason. I wanted to cover why I > think it is both reasonable and important to continue with the vote even > though there are questions about what happens after we decide to > use versions. > > Antoine said: > > Well, the above shows that it's not exactly clear what is being voted > > on, yet. At least it's not clear to me and potentially other bystanders. > > I assume it's clear to you :-) > > I think this is the key issue for the vote, so I want to address it to > explain why I think it's important to continue. > > From my perspective, we spent more than a month discussing versions vs > presets (that is, release by vote with numeric labels vs mechanistic with > time-based labels). To make my perspective easier to see, I'll use an > analogy: we want to make juice and had a long discussion over Apples vs > Oranges. > > That discussion trailed off 2 weeks ago. I had noted that we're really just > trying to decide between Apples and Oranges. We discussed this in the > Parquet sync to see if anyone else (not on the discuss thread) had a strong > opinion. It appeared that we had said all we needed to about the choice of > Apples vs Oranges. > > So I started a vote to affirm what I think is the clear community > preference for Apples (only 2 exceptions). > > Continuing with the analogy a little further, I'm confused because the > objections are (1) "Can you even juice an Apple?" and (2) "Is this a vote > to choose Apples vs Oranges?". For (1), we have examples that this works. > And for (2), I'm struggling to see how this wasn't clear from the context. > The > idea is simply to make a call on the decision point we outlined in our > discussion. (I hope the analogy helps make my perspective easier to > understand.) > > I also think it is important to leave the vote open. I understand the ask: > to write a document on how versions work. But exactly what it means to bump > the minor version of parquet-format is unrelated to our decision to use a > vote to release new format versions. We should focus on the decision point > at hand. In addition, I think it is positive that this has sparked interest > in learning how versioning a format works, so I am reluctant to cancel the > vote only to return to silence. > > Hopefully that makes my thinking clear. Next, how do we move forward? > > Antoine, what if you, Julien, and I work together to finalize Julien's doc > about how versions will work? That way you can ensure the proposal meets > your expectations and I can see that we are making progress. Would that > work for you? > > > > On Thu, Jul 9, 2026 at 11:05 PM Antoine Pitrou <[email protected]> wrote: > > > > > Le 09/07/2026 à 23:00, Ryan Blue a écrit : > > >> The previous thread was entitled "[DISCUSS] Moving to SemVer for > > >> parquet-format releases". :-) > > > > > > We should not conflate parquet-format releases with versions of the > > Parquet > > > spec. Much of the spec is defined in the thrift file of that release, > but > > > it's not the same thing. > > > > I very well hope that we make it the same thing. We have too many > > version numbers already, we don't want to add another one yet. > > > > I certainly assumed that your vote applied to parquet-format releases > > and wasn't proposing a separate versioning space... > > > > >> That's a good point. It deserves encoding somewhere, IMHO. > > > > > > How about a vote? > > > > Well, the above shows that it's not exactly clear what is being voted > > on, yet. At least it's not clear to me and potentially other bystanders. > > I assume it's clear to you :-) > > > > Regards > > > > Antoine. > > > > > > >
