Still thinking through Jan's question; haven't made up my mind on a preference (A, B, C, D, etc.) yet. But had a few questions/reactions, inline.
> My arguments for this approach: 1) Solr has 16993 JIRA issues! > 2) There are 4030 OPEN issues, of which 3681 has not been > touched for a year! Why would we want 3681 OPEN & stale GitHub > issues? Instead I'd like to use StaleBot to tag stale issues+PRs and > auto close+tag if still stale for N days. This is a much-used, proven > approach. To ask what might be a naive question: what's the problem with having many "open but stale" issues, and what does auto-closing things get us? I know it's commonly done, but I've never quite understood why. Folks looking to filter by "active" issues can use the "updated" field that both JIRA and Github provide, so presumably there must be some other benefit? > But if I have a PR (or at least a draft) ready, to me > creating the Github issue is a redundant (and annoying) > copy and paste. Also, I am then confused about where > to discuss and comments (the issue or the PR?). I think this was implicit in Alessandro's message, but to add my 2c more explicitly: I think this can be true regardless of the system for recording PRs or Issues. The duplication is a bit more obvious when both are in Github, but it's going to be there in any setup (including our current one) where issue-reports and code-contributions are separate things. Best, Jason On Wed, Jan 10, 2024 at 5:44 AM Alessandro Benedetti <a.benede...@sease.io> wrote: > Another thing that always bugged me with the Lucene approach is the dualism > issue/PR. > > If I don't have a solution for an issue, it makes complete sense to write > the Github Issue and later on a Pull Request may or may not happen. > > But if I have a PR (or at least a draft) ready, to me creating the Github > issue is a redundant (and annoying) copy and paste. > Also, I am then confused about where to discuss and comments (the issue or > the PR?). > Effectively nowadays Github PRs have plenty of description, discussion and > review capabilities so in case it's a bug-fix or a well-established code > direction, does it make sense to have the associated issue? > > I understand that ideally, you would like to have an issue first, discuss > it with other committers and then start the implementation, but being > honest I realised over the years that this makes contributing a full-time > job and most of us(who are not lucky enough to be paid to contribute) can't > (and shouldn't) commit to that. > So to me, it happens all the time I go deep with a bug fix or new feature, > open the PR and then open the Jira issue. > > Could we possibly go with Solr in a direction where there's always at least > one PR for an issue, but not necessarily an issue for a PR? > Just food for thought based on my experience, > Cheers > -------------------------- > *Alessandro Benedetti* > Director @ Sease Ltd. > *Apache Lucene/Solr Committer* > *Apache Solr PMC Member* > > e-mail: a.benede...@sease.io > > > *Sease* - Information Retrieval Applied > Consulting | Training | Open Source > > Website: Sease.io <http://sease.io/> > LinkedIn <https://linkedin.com/company/sease-ltd> | Twitter > <https://twitter.com/seaseltd> | Youtube > <https://www.youtube.com/channel/UCDx86ZKLYNpI3gzMercM7BQ> | Github > <https://github.com/seaseltd> > > > On Wed, 10 Jan 2024 at 10:26, Alessandro Benedetti <a.benede...@sease.io> > wrote: > > > Hi all, > > thanks for raising this! > > > > I am in favour of: > > B) Migrate only OPEN JIRA issues, refer to JIRA for ancient history > > > > But rather than just OPEN, I would probably migrate only the active ones? > > > > I agree it doesn't make sense to duplicate thousands of Jiras. > > > > I would also be ok with C, mine is just a preference not a veto at all. > > Cheers > > -------------------------- > > *Alessandro Benedetti* > > Director @ Sease Ltd. > > *Apache Lucene/Solr Committer* > > *Apache Solr PMC Member* > > > > e-mail: a.benede...@sease.io > > > > > > *Sease* - Information Retrieval Applied > > Consulting | Training | Open Source > > > > Website: Sease.io <http://sease.io/> > > LinkedIn <https://linkedin.com/company/sease-ltd> | Twitter > > <https://twitter.com/seaseltd> | Youtube > > <https://www.youtube.com/channel/UCDx86ZKLYNpI3gzMercM7BQ> | Github > > <https://github.com/seaseltd> > > > > > > On Mon, 8 Jan 2024 at 23:57, Jan Høydahl <jan....@cominvent.com> wrote: > > > >> Bringing attention to this thread again. > >> > >> Now that Lucene has some experience after the migration and with using > >> Issues, labels etc, I'd like to discuss whether we'd want to copy the > >> Lucene approach or do something different. > >> > >> I'm not frequenting the Lucene issue tracker much, and am not either > >> aware of a formal evaluation of their issue migration. So appreciate > >> feedback from committers who have more exposure from Lucene. > >> > >> In my mind, we, Solr, have four options > >> A) Migrate everything, like Lucene did > >> B) Migrate only OPEN JIRA issues, refer to JIRA for ancient history > >> C) Fresh-start empty GH issues, use JIRA as archive before yyyy-mm-dd > >> D) Continue with g'old JIRA > >> > >> I was getting used to the thought of copying Lucene's approach, but > >> re-thinking now, I have once again shifted my preference towards C), a > >> fresh start. I'll summarize my reasons below with some numbers and > >> experience from Lucene's GH issues. Forgive my last rant on this > subject :) > >> > >> <rant> > >> I'm a fan of NOT migrating the entire JIRA history to GH. Instead let > the > >> R/O SOLR-JIRA be the system of record for historic issues forever. I.e. > >> start a fresh, empty GH issue tracker for all NEW issues. The main con, > two > >> systems of record, can imo be mitigated with SearchEngineTechnoloty™. > >> Nothing is lost and the duplication of 16k issues in two systems is more > >> confusing imo. We could time-box the overlap period where both systems > are > >> writable to, say 3 months, and at the end of that period, CLOSE all > >> still-open JIRA issues with a label and a suitable message. > >> > >> My arguments for this approach: 1) Solr has 16993 JIRA issues! 2) There > >> are 4030 OPEN issues, of which 3681 has not been touched for a year! Why > >> would we want 3681 OPEN & stale GitHub issues? Instead I'd like to use > >> StaleBot to tag stale issues+PRs and auto close+tag if still stale for N > >> days. This is a much-used, proven approach. 3) The Lucene migration > caused > >> a whopping 642 issue/PR labels < > >> https://github.com/apache/lucene/issues/labels>, impossible to browse > >> manually. Most labels are trying to record legacy JIRA fields. I checked > >> e.g. the "affects-version" < > >> https://github.com/apache/lucene/labels?q=affects-version:9>, label in > >> Lucene. New labels have not been maintained for recent releases (8.11.2, > >> 9.4..9), and those labels are ~NEVER even used when people create new GH > >> issues, so why even bother? Same for Priority etc. > >> </rant> > >> > >> Let the discussion continue... > >> > >> Jan > >> > >> > >> > 3. nov. 2023 kl. 12:33 skrev Jan Høydahl <jan....@cominvent.com>: > >> > > >> > Bringing this to our attention again. Lucene seems to have survived > >> well after the migration to Github issues. They have established a way > to > >> work with milestones (https://github.com/apache/lucene/milestones) and > >> labels (https://github.com/apache/lucene/labels?q=type), and they have > >> updated release-wizard with corresponding steps. > >> > > >> > So are we ready to follow their lead? > >> > > >> > Jan > >> > > >> >> 18. okt. 2022 kl. 12:58 skrev Jan Høydahl <jan....@cominvent.com>: > >> >> > >> >> +1 from me too. > >> >> > >> >> I'm still not sold on bringing all history from JIRA into GH but > >> that's what Lucene did, so perhaps just doing the same (+ lessons > learnt) > >> is the smoothest path. > >> >> Solr would in addition need to find a new process for security > issues. > >> But we could just fall back on plain security@solr mailing list until a > >> new solution is ready. > >> >> > >> >> Jan > >> >> > >> >>> 17. okt. 2022 kl. 16:20 skrev David Smiley <dsmi...@apache.org>: > >> >>> > >> >>> +1 to migrate. > >> >>> > >> >>> Yeah. Maybe Tomoko could validate the steps required? (CC'ed) Jeb > >> listed > >> >>> them in JIRA; the steps/mechanics can be discussed there while we > >> leave > >> >>> this thread as voting on the major decision. > >> >>> > >> >>> ~ David Smiley > >> >>> Apache Lucene/Solr Search Developer > >> >>> http://www.linkedin.com/in/davidwsmiley > >> >>> > >> >>> > >> >>> On Mon, Oct 17, 2022 at 10:12 AM Houston Putman <hous...@apache.org > > > >> wrote: > >> >>> > >> >>>> I'm a big +1 on this idea, just like I was for Lucene's migration. > >> >>>> > >> >>>> Also I think that we could very much mooch off of the monumental > >> amounts of > >> >>>> hard work that Tomoko and Mike did for Lucene. > >> >>>> > >> >>>> There would certainly still be manual work, and changes to their > >> script > >> >>>> needed, but I don't think it would be as back-breaking of a task. > >> >>>> > >> >>>> - Houston > >> >>>> > >> >>>> On Fri, Oct 14, 2022 at 1:07 AM Noble Paul <noble.p...@gmail.com> > >> wrote: > >> >>>> > >> >>>>> I agree that JIRA is one extra step that is not adding a lot of > >> value. > >> >>>>> Github issues are definitely better > >> >>>>> > >> >>>>> On Fri, Oct 14, 2022 at 3:04 PM David Smiley <dsmi...@apache.org> > >> wrote: > >> >>>>> > >> >>>>>> Sharing for visibility. > >> >>>>>> > >> >>>>>> ~ David Smiley > >> >>>>>> Apache Lucene/Solr Search Developer > >> >>>>>> http://www.linkedin.com/in/davidwsmiley > >> >>>>>> > >> >>>>>> > >> >>>>>> ---------- Forwarded message --------- > >> >>>>>> From: Jeb Nix (Jira) <j...@apache.org> > >> >>>>>> Date: Mon, Oct 10, 2022 at 7:11 PM > >> >>>>>> Subject: [jira] [Created] (SOLR-16455) Migrate Jira to Github > >> Issues > >> >>>> and > >> >>>>>> Github Projects, and migrate mailing lists to Github Discussions > >> >>>>>> To: <iss...@solr.apache.org> > >> >>>>>> > >> >>>>>> > >> >>>>>> Jeb Nix created SOLR-16455: > >> >>>>>> ------------------------------ > >> >>>>>> > >> >>>>>> Summary: Migrate Jira to Github Issues and Github > >> >>>> Projects, > >> >>>>>> and migrate mailing lists to Github Discussions > >> >>>>>> Key: SOLR-16455 > >> >>>>>> URL: > >> https://issues.apache.org/jira/browse/SOLR-16455 > >> >>>>>> Project: Solr > >> >>>>>> Issue Type: Wish > >> >>>>>> Security Level: Public (Default Security Level. Issues are > >> >>>> Public) > >> >>>>>> Components: github > >> >>>>>> Reporter: Jeb Nix > >> >>>>>> > >> >>>>>> > >> >>>>>> GitHub is where people are at when they lookup for Solr (or > >> basically > >> >>>> any > >> >>>>>> project). Most of the modern projects that have been started with > >> Jira > >> >>>>> and > >> >>>>>> mailing lists have migrated to Github in the last few years. > >> Lucene did > >> >>>>>> that just now for the Issues which has allowed me to explore much > >> more > >> >>>> of > >> >>>>>> their issues. GitHub works great and many think that it works > even > >> >>>> better > >> >>>>>> (I think that there is no doubt that it is working better for the > >> >>>>>> Discussions vs. Mailing lists). > >> >>>>>> > >> >>>>>> I suggest here a pretty heavy move, that personally will allow me > >> to > >> >>>>> start > >> >>>>>> anticipating within Solr's community (since I really don't like > the > >> >>>>> mailing > >> >>>>>> lists nor Jira), and I think that there are much more like me out > >> >>>> there. > >> >>>>> In > >> >>>>>> my opinion, when the issues are managed on Github, it is much > >> simpler > >> >>>> to > >> >>>>>> collaborate and they will get wider exposure since developers are > >> >>>>> spending > >> >>>>>> time on Github anyway (whether if it's for their projects or for > >> >>>> looking > >> >>>>> at > >> >>>>>> the actual source code). It is also important to mention that it > is > >> >>>>> pretty > >> >>>>>> cumbersome for a new contributor that wants to add stuff to Solr, > >> to > >> >>>> talk > >> >>>>>> about this via mail, then translate them to Jira of the issues, > and > >> >>>> just > >> >>>>>> after that submit a PR on Github. e.g. 3 different systems for > each > >> >>>>>> process. > >> >>>>>> > >> >>>>>> Actually, I thought such a great move (for me at least) would > never > >> >>>>> happen > >> >>>>>> in Solr in the next years since I didn't think that the community > >> sees > >> >>>> & > >> >>>>>> understands the many advantages yet. But now that the Lucene guys > >> did > >> >>>>> this, > >> >>>>>> I believe that it is possible for Solr too. > >> >>>>>> > >> >>>>>> > >> >>>>>> > >> >>>>>> -- > >> >>>>>> This message was sent by Atlassian Jira > >> >>>>>> (v8.20.10#820010) > >> >>>>>> > >> >>>>>> > >> --------------------------------------------------------------------- > >> >>>>>> To unsubscribe, e-mail: issues-unsubscr...@solr.apache.org > >> >>>>>> For additional commands, e-mail: issues-h...@solr.apache.org > >> >>>>>> > >> >>>>> > >> >>>>> > >> >>>>> -- > >> >>>>> ----------------------------------------------------- > >> >>>>> Noble Paul > >> >>>>> > >> >>>> > >> >> > >> > > >> > >> >
