Oh, sorry for the misunderstanding: I didn't mean to suggest to actually use confidence intervals, I added that information just as a curiosity. I agree that we should make the decision based on the explicit % of votes.
On Tue, Dec 12, 2017 at 4:17 PM Kenneth Knowles <[email protected]> wrote: > -user list for this > > FWIW confidence intervals were not mentioned before the poll was launched, > but in fact it was pretty explicitly just votes <= 5%. > > If we require that the 95% confidence interval falls under 5% we either > need 2000 respondents (not going to happen) at current proportions or for a > more reasonable 300 respondents the actual proportion would be ~2.3% which > is 7 votes, while we already have 9 votes. In other words, the poll is > already effectively done if that is the standard. > > But I would favor tilting the bias the other way. I'm not just trying to > make the data fit the desired conclusion. If the original proposal had > talked about confidence intervals, or the implied result that we keep Java > 7 support for 3% of responses (out of expected ~500 or less) I would have > said the balance was off. > > Kenn > > On Mon, Dec 11, 2017 at 9:00 AM, Eugene Kirpichov <[email protected]> > wrote: > >> Intermediate update: right now it's 219 votes on Twitter (and 0 on this >> thread: seems that people find Twitter a more convenient medium!), with 96% >> who are either on Java 8 or can easily switch, vs. 4% who can not easily >> switch. >> >> Using a proportion confidence interval calculator, so far we're looking >> at between 1.9 and 7.7% of users (at 95% confidence) needing Java 7 support >> :) >> >> Willing users can help gather more votes by RT'ing the poll or chiming in >> on this thread! >> >> So: for now we don't yet have enough confidence that Java 7 support can >> be dropped from the SDK (to remind, we'll conclude the vote on January 7th; >> since people so overwhelmingly vote on Twitter, we may consider doing 3 >> more rounds of the same poll?..), however: >> - It may be enough confidence to resolve to *build* the Beam SDK using >> JDK8 (at Java 7 source language level), which is something +Ismaël Mejía >> <[email protected]> and +Daniel Oliveira <[email protected]> once >> requested as part of building Java 9 support. I suspect that the sets >> "users who need to build their own Beam SDK" and "users who can not upgrade >> to Java 8" have a vanishingly small intersection. >> - Then, we may consider also start building *tests* at Java 8 language >> level, which will give the bulk of the benefit of encouraging >> Java8-friendly APIs. >> Thoughts on the above? >> >> On Thu, Dec 7, 2017 at 6:22 PM Eugene Kirpichov <[email protected]> >> wrote: >> >>> A Twitter poll has been sent out too >>> https://twitter.com/ApacheBeam/status/938926195910905857 >>> However, due to limitations of Twitter it can only be open for 7 days. I >>> encourage people who are late to the poll to comment on this thread instead. >>> >>> On Thu, Dec 7, 2017 at 3:50 PM Eugene Kirpichov <[email protected]> >>> wrote: >>> >>>> Followup: we would like to keep this poll open for 2 weeks, however >>>> some people have expressed concern that this is too short. >>>> Let us keep this poll open for 1 month starting today. So far the >>>> agreed-upon decision has been to move forward with the plan if fewer than >>>> 5% of all respondents choose option 3. >>>> >>>> On Thu, Dec 7, 2017 at 3:48 PM Eugene Kirpichov <[email protected]> >>>> wrote: >>>> >>>>> This is a follow-up on a previous similar thread >>>>> https://lists.apache.org/thread.html/2e1890c62d9f022f09b20e9f12f130fe9f1042e391979087f725d2e0@%3Cuser.beam.apache.org%3E >>>>> in >>>>> which the community consistently expressed support for transitioning Beam >>>>> Java to Java8-only. >>>>> >>>>> Now that the release of Beam 2.2.0 has completed, we are considering >>>>> performing this change specifically in the immediate next release: Beam >>>>> 2.3.0, i.e. dropping support for Java 7 without a bump in the major >>>>> version >>>>> of Beam. >>>>> >>>>> The reasons for switching to Java 8 in general are considered in the >>>>> thread above. >>>>> The reasons in favor of making the switch in Beam 2.3.0 are as follows: >>>>> - It is believed that usage of Java 7 in production is already >>>>> vanishingly small. >>>>> - Since Java 7 has not been receiving even security updates for years, >>>>> helping perpetuate its usage would be a bad idea >>>>> - A major version bump is a major step and would likely happen only >>>>> after a large number of other major changes in Beam accumulate - i.e. many >>>>> months. Maintaining Java 7 compatibility for that long would have costs, >>>>> including the awkward possibility of switching Beam to Java 8 after Java >>>>> 8's end of life (September 2018 AFAIK) >>>>> - Updating to Java 8 would lead to Beam more quickly gaining more >>>>> Java8-friendly APIs, because Beam SDK authors and contributors would have >>>>> more liberty, more responsibility and more experience with working in the >>>>> context of Java8. Delaying until Beam 3.0 would delay this as well. >>>>> >>>>> With that in mind, we'd like to poll the Beam community to gather >>>>> information about usage of Java 7 and Java 8 in production. Please vote: >>>>> >>>>> Option 1. I am already using only Java 8+ for building my production >>>>> Beam code. >>>>> >>>>> Option 2. I am using Java 7 for building my production Beam code, but >>>>> I would have no trouble with the switch to Java 8 [e.g. my transition to >>>>> Java 8 would be easy and/or I don't expect that I'll have strong reasons >>>>> to >>>>> upgrade to Beam 2.3 anyway]. >>>>> >>>>> Option 3. I am using Java 7 for building my production Beam code, and >>>>> dropping Java 7 would be a blocker or hindrance to adopting the new >>>>> release >>>>> for me [e.g. I expect that I'll have strong reasons to update to Beam 2.3, >>>>> but I expect that it will be difficult because of lack of Java 7 support] >>>>> >>>>> >
