Hi Kenn,yes, I can do that. I think that there should be at least one first BIP, I can try to setup one. But (as opposed to my previous proposal) I'll try to setup a fresh one, not the one of [BEAM-8550], because that one already has a PR and rebasing the PR on master for such a long period (and it is likely, that final polishing of the BIP process will take several more months) starts to be costly. I have in mind two fresh candidates, so I'll pick one of them. I think that only setuping a cwiki would not start the process, we need a real-life example of a BIP included in that.
Does that sound ok? Jan On 2/1/20 5:55 AM, Kenneth Knowles wrote:
These stages sound like a great starting point to me. Would you be the volunteer to set up a cwiki page for BIPs?KennOn Mon, Jan 20, 2020 at 3:30 AM Jan Lukavský <je...@seznam.cz <mailto:je...@seznam.cz>> wrote:I agree that we can take inspiration from other projects. Besides the organizational part (what should be part of BIP, where to store it, how to edit it and how to make it available to the whole community) - for which the cwiki might be the best option - there are still open questions about the lifecycle of a BIP: a) when to create one? - I feel this might be optional, there might be some upper bound of features that are really "too big" or "too controversial", so these should undergo the BIP process in all cases, otherwise the decision might be part of the proposal, the question is how to define those b) what are lifecycle stages of a BIP? How to do transitions between these stages? - From the top of my head this might be: a) proposal -- not yet accepted b) accepted -- most probably after vote? c) in progress -- having assigned JIRA and being worked on d) done -- after merge to master e) released -- obvious WDYT? Jan On 1/15/20 8:19 PM, Kenneth Knowles wrote:Focusing this thread on the BIP process seems wise, without changing much else in the same thread. I don't think the BIP process has to do with exactly how design docs are written or archived, but the ability to *at a glance* understand: - what are the high level proposals - status of the proposals - who to contact - how to get to more info (links to design docs, thread, Jiras, etc) A page with a table on cwiki is common and seems good for this. How we manage such a table would be a possible next step. I think they should focus on large changes that need heavyweight process, so should encourage lightweight creation. I think adding heavy process to smaller changes would be bad. ---- I have looked multiple times at other projects (linked in prior thread and in this thread too but gathering them here) Spark: https://spark.apache.org/improvement-proposals.html - Jira is not good for "at a glance" reading. The title should have a short and easy to understand paragraph. Kafka: https://cwiki.apache.org/confluence/display/KAFKA/Kafka+Improvement+Proposals - Quite a lot of content; I would prefer 10s of proposals. But it is readable enough. Table lacks important content like links and summaries. - Blends the table with a bunch of header material that IMO ets in the way Flink: https://cwiki.apache.org/confluence/display/FLINK/Flink+Improvement+Proposals - Looks very similar to Kafka - Target Release is too specific, and actual status is missing Airflow: https://cwiki.apache.org/confluence/display/AIRFLOW/Airflow+Improvements+Proposals - seems best organized, and the table has more info - having sections for the different status proposals in different tables is great - "InRelease" column is left blank It seems there is a lot of redundancy with Jira fields - owner, release, etc. I think that redundancy is good. If it is too much effort to redundantly manage to write it in the table then it probably is not appropriate for heavyweight process. Anything that is one simple task that fits in a Jira that can be passed around from person to person shouldn't be a BIP. Probably anything where we can guess the target version isn't big enough for a BIP. Kenn On Thu, Jan 9, 2020 at 7:59 AM Jan Lukavský <je...@seznam.cz <mailto:je...@seznam.cz>> wrote: I think that, besides ownership of a feature, a BIP (or whatever document or process) should contain the following: * description of a problem that the improvement addresses - this is currently often part of design doc * description of multiple possible solutions (if multiple exist, which is probably mostly the case) * justifying choice of a particular solution * result of a vote - the vote should cover both (a) do we don't this feature in the first place and (b) do we accept the proposed solution This would probably be iterative process involving multiple people, mailing list communication, etc. Pretty much what we do now, just there would be a place to keep track of decisions made throughout the process. I pretty much think that voting on complicated solutions is vital, the soft consensus approach is good for "simple" features (what that means might be subjective), but might fail for features where multiple more or less complex solutions exist. After successful PMC vote, the problem simplifies to reviewing code, the reviewer doesn't have to think about "do we want this feature?". That is given in advance. After we agree on the process and the form it should have I can volunteer to test it by letting proposal of ordered stateful processing pass through it. On 1/9/20 9:11 AM, Alex Van Boxel wrote:Maybe tweaking the current process a bit is enough. I like the Docs for having discussions but there no good as a /proper design document/, for the following reasons: I see design documents full of discussions and wonder: * Who will be the *main owner* and the *co-owners* (meaning people that are invested of bringing this forward and can *act* as /reviewers/). I think a proposal needs especially this: ownership * Lack of visibility of final state? Or is it superseded by another proposal. A final state could include the votes... * Does the proposal need amendments. An example, while implementing the proposal, we see that something in the design was lacking and needs to be added. So the Docs are great, but maybe we should a few mandatory blocks and a few rules: * *Resolve all discussions* before switching to final state. * If new discussions pop up, maybe an amendment needs to be made (or correct). Corrections could be added to a *changelog* in the beginning. * If a new proposal supersedes on, both should be linked * Most importantly: Who can act as *owner* end reviewers for this proposal. _/ _/ Alex Van Boxel On Thu, Jan 9, 2020 at 7:59 AM Kenneth Knowles <k...@apache.org <mailto:k...@apache.org>> wrote: It does seem that the community would find this useful. I agree with Robert that it has downsides and it is not appropriate all the time. We added https://beam.apache.org/roadmap/ a little while ago. I think that the granularity of a BIP is about the same as the granularity of what we would want to show to users on a roadmap on our public site. So we sort of already have this. Perhaps we want to formalize changes to the roadmap and only include voted upon approved BIPs on the roadmap on the web site. The current roadmap should be viewed as a crowd sourced bootstrap, for sure. Imagine a roadmap that a company shares with a customer. The most important thing is to be extremely clear about what is intended to be built, when it is expected, and how they can follow the developments. And for the open source community, it should be clear what they can expect to work on and know that the project / PMC has agreed on the feature and will not push back after some effort has been put into it. Kenn On Tue, Dec 17, 2019 at 11:07 AM Jan Lukavský <je...@seznam.cz <mailto:je...@seznam.cz>> wrote: Hi, I feel a "soft consensus" :) that people see some benefits of introducing (possibly optional) process of proposing new features. I think that in order to proceed with this we need to agree on goals that we want to achieve. Whether the process should or should not be optional, which form it should have, and answers on all these other questions could be answered after that. So, I'll try to state some issues I see with our current approach, please feel free to correct any of them, or add any other: - due to the "soft consensus" approach, we actually delegate the final responsibility of "feature acceptance" to reviewer(s) - these might or might not be happy with that - by splitting this into first-consensus-then-implementation-then-review approach, we remove the burden of responsibility of respective feature from reviewers - they can focus only on the main purpose of the review - that is verifying the quality of code - as mentioned before, this brings better visibility to (core) features - and last but not least makes it possible to prioritize work and build more complex long-term goals I think it is essential to have a consensus on whether or not these are some points we want to target (that is, we see our current approach as sub-optimal in these areas) or not. Jan On 12/17/19 7:08 PM, Pablo Estrada wrote:It seems that lots of people see benefit in a more formalized BIP process. I think that makes sense, though I'd like to give people the freedom to choose the medium for their design discussions. The projects I'm aware of usually do this through wiki-type mediums. We have cwiki, though lots of people like working with Gdocs' collaboration features. Are there other mediums that could be used for this? A possible implementation is: We could keep cwiki as the 'index' - so anyone proposing a new BIP would have to add a new BIP entry in the cwiki, but they'd be free to link to a Gdoc from there, or to develop the proposal in the cwiki entry itself. Thoughts? Best -P. On Tue, Dec 17, 2019 at 9:14 AM Maximilian Michels <m...@apache.org <mailto:m...@apache.org>> wrote: The main benefit of BIPs I see is the visibility they create for the project users and contributors. Right now, we have a long unordnered list of design documents. Some of the documents are not even in that list. With BIPs, we would end up with an ordered list "BIP-1, BIP-2, .." which reflects important design decisions over time. Simply assigning an id, makes it a lot more formal. In my eyes, the id assignment would also require that you communicate the changes in a way that the community can accept the proposal, preferably via lazy consensus. All in all, this could help communicate changes in Beam better. JIRA, on the other hand, contains concrete implementation steps and all kinds of other changes. Cheers, Max On 16.12.19 21:41, Robert Bradshaw wrote: > Additional process is a two-edged sword: it can help move stuff > forward, to the correct decision, but it can also add significant > overhead. > > I think there are many proposals for which the existing processes of > deriving consensus (over email, possibly followed by a formal vote or > lazy consensus) are sufficient. However, sometimes they're not. > Specifically, for long-term roadmaps, it would be useful to have them > in a standard place that can be tracked and understood (I don't think > we've been able to use JIRA effectively for this here). I also think > there are some proposals that reach a certain level of complexity that > trying to address them by occasionally responding to email threads as > they come up is insufficient. For these latter, I think there is a > need for commitment for a group of people in the community to commit > to clearly defining and driving a solution to the problem via a more > formal process. Often the one making the proposal has sufficient > motivation, but sometimes what lacks is be (non-sporadic) investment > by those trying to understand, evaluate, and incorporate the proposal. > > So I'm (strongly) +1 for exploring a more formal process, but -1 on > requiring it. > > On Sun, Dec 15, 2019 at 1:07 AM Jan Lukavský <je...@seznam.cz <mailto:je...@seznam.cz>> wrote: >> >> Hi, >> >> thanks for reactions so far. I agree that there are many questions that have to be clarified. I'd propose to split this into two parts: >> >> a) first reach a consensus that we want this process in the first place >> >> b) after that, we need to clarify all the details - that will probably be somewhat iterative procedure >> >> I'm not sure if there is something more we need to clarify before we can cast a vote on (a). >> >> Thoughts? >> >> Jan >> >> On 12/10/19 3:46 PM, Łukasz Gajowy wrote: >> >> +1 for formalizing the process, enhancing it and documenting clearly. >> >> I noticed that Apache Airflow has a cool way of both creating AIPs and keeping track of all of them. There is a "Create new AIP" button on their Confluence. This way, no AIP gets lost and all are kept in one place. Please keep in mind that this is also the problem we want to solve in Beam and try to keep track of all the documents we have so far*. It's certainly good to solve that problem too, if possible. >> >> Also the AIP structure is something that I find nice - There's place for all additional resources, JIRAs, discussion in comments and state of the proposal. Even if we don't choose to use Confluence, we definitely could use a similar template with all that information for our google docs proposals or any other tool we stick to. >> >> Thanks! >> >> *thank you, Ismael and Alexey, for all the reminders under the proposals to add them to Confluence list! :) >> >> wt., 10 gru 2019 o 13:29 jincheng sun <sunjincheng...@gmail.com <mailto:sunjincheng...@gmail.com>> napisał(a): >>> >>> Thanks for bring up this discussion Jan! >>> >>> +1 for cearly define BIP for beam. >>> >>> And I think would be nice to initialize a concept document for BIP. Just a reminder: the document may contains: >>> >>> - How many kinds of improvement in beam. >>> - What kind of improvement should to create a BIP. >>> - What should be included in a BIP. >>> - Who can create the BIP. >>> - Who can participate in the discussion of BIP and who can vote for BIP. >>> - What are the possible limitations of BiP, such as whether it is necessary to complete the dev of BIP in one release. >>> - How to track a BIP. >>> >>> Here is a question: I found out a policy[1] in beam, but only contains the poilcy of release , my question is does beam have something called Bylaws? Similar as Flink[1]. >>> >>> Anyway, I like your proposals Jan :) >>> >>> Best, >>> Jincheng >>> [1] https://beam.apache.org/community/policies/ >>> [2] https://cwiki.apache.org/confluence/display/FLINK/Flink+Bylaws#FlinkBylaws-Approvals >>> >>> >>> David Morávek <david.mora...@gmail.com <mailto:david.mora...@gmail.com>> 于2019年12月10日周二 下午2:33写道: >>>> >>>> Hi Jan, >>>> >>>> I think this is more pretty much what we currently do, just a little bit more transparent for the community. If the process is standardized, it can open doors for bigger contributions from people not familiar with the process. Also it's way easier to track progress of BIPs, than documents linked from the mailing list. >>>> >>>> Big +1 ;) >>>> >>>> D. >>>> >>>> On Sun, Dec 8, 2019 at 12:42 PM Jan Lukavský <je...@seznam.cz <mailto:je...@seznam.cz>> wrote: >>>>> >>>>> Hi, >>>>> >>>>> I'd like to revive a discussion that was taken some year and a half ago >>>>> [1], which included a concept of "BIP" (Beam Improvement Proposal) - an >>>>> equivalent of "FLIP" (flink), "KIP" (kafka), "SPIP" (spark), and so on. >>>>> >>>>> The discussion then ended without any (public) conclusion, so I'd like >>>>> to pick up from there. There were questions related to: >>>>> >>>>> a) how does the concept of BIP differ from simple plain JIRA? >>>>> >>>>> b) what does it bring to the community? >>>>> >>>>> I'd like to outline my point of view on both of these aspects (they are >>>>> related). >>>>> >>>>> BIP differs from JIRA by definition of a process: >>>>> >>>>> BIP -> vote -> consensus -> JIRA -> implementation >>>>> >>>>> This process (although it might seem a little unnecessary formal) brings >>>>> the following benefits: >>>>> >>>>> i) improves community's overall awareness of planned and in-progress >>>>> features >>>>> >>>>> ii) makes it possible to prioritize long-term goals (create "roadmap" >>>>> that was mentioned in the referred thread) >>>>> >>>>> iii) by casting explicit vote on each improvement proposal diminishes >>>>> the probability of wasted work - as opposed to our current state, where >>>>> it is hard to tell when there is a consensus and what actions need to be >>>>> done in order to reach one if there isn't >>>>> >>>>> iv) BIPs that eventually pass a vote can be regarded as "to be >>>>> included in some short term" and so new BIPs can build upon them, >>>>> without the risk of having to be redefined if their dependency for >>>>> whatever reason don't make it to the implementation >>>>> >>>>> Although this "process" might look rigid and corporate, it actually >>>>> brings better transparency and overall community health. This is >>>>> especially important as the community grows and becomes more and more >>>>> distributed. There are many, many open questions in this proposal that >>>>> need to be clarified, my current intent is to grab a grasp about how the >>>>> community feels about this. >>>>> >>>>> Looking forward to any comments, >>>>> >>>>> Jan >>>>> >>>>> [1] >>>>> https://lists.apache.org/thread.html/4e1fffa2fde8e750c6d769bf4335853ad05b360b8bd248ad119cc185%40%3Cdev.beam.apache.org%3E >>>>>