You can also do the trick of having two branches that diverge at the very first commit. C++ and Java are in the same repository but if you have a given branch checked out you are looking at one or the other.
That said, I’d be inclined to put everything in the same repo. The repo size will not be prohibitive as long as we follow Julien’s recommendation to put large objects (e.g. test data sets) elsewhere. And it makes it possible for a single patch to update both C++ and Java code lines, and also update shared content (the specification and the source files for the web site). And having said THAT, it doesn’t really matter. If we realize we’ve made a horrible mistake in 2 years we can split the repo or merge the repos. Julian > On Dec 17, 2015, at 11:09 AM, Julien Le Dem <[email protected]> wrote: > > I guess what I meant is: separate repos => separate releases. > One repo can still have separate releases as was mentioned. > > I’m not too fond of the separate repos in parquet anymore. > The only reason I would split to a separate repo now is if we have large > files for regression testing and that’s because of how git works. > > Separate directories java and cpp seems a given since they will build > independently. > > I’d vote for just one repo with /format /java /cpp. > both /java and /cpp depend on /format > > releasing the format independently is useful because the semantics are > different than for an API. > A breaking change in the format may not be an API change and vice-versa. > > Julien > > >> On Dec 16, 2015, at 7:03 PM, Ted Dunning <[email protected]> wrote: >> >> >> One repo should be a given. >> >> Separate directories should be the question. >> >> >> >> On Wed, Dec 16, 2015 at 6:21 PM, Jason Altekruse <[email protected] >> <mailto:[email protected]>> wrote: >> I think that is a worthwhile discussion, parallel vs independent releases, >> but I don't understand why it relates to one repo or not. Couldn't the >> release tag names just include the language (cpp, java python)? What other >> parts of version control are related to releasing? >> >> On Wed, Dec 16, 2015 at 5:57 PM, Julien Le Dem <[email protected] >> <mailto:[email protected]>> wrote: >> for the git repos it boils down to wether we want to release arrow-cpp and >> arrow-java independently or together with the same version numbers. >> >> >> >> On Tue, Dec 15, 2015 at 7:15 PM, Jacques Nadeau <[email protected] >> <mailto:[email protected]>> wrote: >> Thanks Wes, that's great! >> On Dec 14, 2015 9:44 AM, "Wes McKinney" <[email protected] >> <mailto:[email protected]>> wrote: >> >>> hi folks, >>> >>> In the interim I created a new public GitHub organization to host code >>> for this effort so we can organize ourselves in advance of more >>> progress in the ASF: >>> >>> https://github.com/arrow-data <https://github.com/arrow-data> >>> >>> I have a partial C++ implementation of the Arrow spec that I can move >>> there, along with a to-be-Markdown-ified version of a specification >>> subject to more iteration. The more pressing short term matter will be >>> making some progress on the metadata / data headers / IPC protocol >>> (e.g. using Flatbuffers or the like). >>> >>> Thoughts on git repo structure? >>> >>> 1) Avro-style — "one repo to rule them all" >>> 2) Parquet-style — arrow-format, arrow-cpp, arrow-java, etc. >>> >>> (I'm personally more in the latter camp, though integration tests may >>> be more tedious that way) >>> >>> Thanks >>> >>> On Thu, Dec 3, 2015 at 4:18 PM, Jacques Nadeau <[email protected] >>> <mailto:[email protected]>> wrote: >>>> I've opened a name search for our top vote getter. >>>> >>>> https://issues.apache.org/jira/browse/PODLINGNAMESEARCH-92 >>>> <https://issues.apache.org/jira/browse/PODLINGNAMESEARCH-92> >>>> >>>> >>>> I also just realized that my previously email dropped other recipients. >>>> Here it is below. >>>> >>>> ---- >>>> I think we can call the voting closed. Top vote getters: >>>> >>>> Apache Arrow (17) >>>> Apache Herringbone (9) >>>> Apache Joist (8) >>>> Apache Colbuf (8) >>>> >>>> I'll up a PODLINGNAMESEARCH-* shortly for Arrow. >>>> >>>> --- >>>> >>>> >>>> >>>> >>>> >>>> >>>> -- >>>> Jacques Nadeau >>>> CTO and Co-Founder, Dremio >>>> >>>> On Thu, Dec 3, 2015 at 1:23 AM, Marcel Kornacker <[email protected] >>>> <mailto:[email protected]>> >>>> wrote: >>>>> >>>>> Just added my vote. >>>>> >>>>> On Thu, Dec 3, 2015 at 12:51 PM, Wes McKinney <[email protected] >>>>> <mailto:[email protected]>> wrote: >>>>>> Shall we call the voting closed? Any last stragglers? >>>>>> >>>>>> On Tue, Dec 1, 2015 at 5:39 PM, Ted Dunning <[email protected] >>>>>> <mailto:[email protected]>> >>>>>> wrote: >>>>>>> >>>>>>> Apache can handle this if we set the groundwork in place. >>>>>>> >>>>>>> Also, Twitter's lawyers work for Twitter, not for Apache. As such, >>>>>>> their >>>>>>> opinions can't be taken by Apache as legal advice. There are issues >>> of >>>>>>> privilege, conflict of interest and so on. >>>>>>> >>>>>>> >>>>>>> >>>>>>> On Wed, Dec 2, 2015 at 7:51 AM, Alex Levenson >>>>>>> <[email protected] <mailto:[email protected]>> >>>>>>> wrote: >>>>>>>> >>>>>>>> I can ask about whether Twitter's lawyers can help out -- is that >>>>>>>> something we need? Or is that something apache helps out with in the >>>>>>>> next >>>>>>>> step? >>>>>>>> >>>>>>>> On Mon, Nov 30, 2015 at 9:32 PM, Julian Hyde <[email protected] >>>>>>>> <mailto:[email protected]>> >>> wrote: >>>>>>>>> >>>>>>>>> +1 to have a vote tomorrow. >>>>>>>>> >>>>>>>>> Assuming that Vector is out of play, I just did a quick search for >>>>>>>>> the >>>>>>>>> top 4 remaining, (“arrow”, “honeycomb”, “herringbone”, “joist"), at >>>>>>>>> sourceforge, open hub, trademarkia, and on google. There are no >>>>>>>>> trademarks >>>>>>>>> for these in similar subject areas. There is a moderately active >>>>>>>>> project >>>>>>>>> called “joist” [1]. >>>>>>>>> >>>>>>>>> I will point out that “Apache Arrow” has native-american >>> connotations >>>>>>>>> that we may or may not want to live with (just ask the Washington >>>>>>>>> Redskins >>>>>>>>> how they feel about their name). >>>>>>>>> >>>>>>>>> If someone would like to vet other names, use the links on >>>>>>>>> https://issues.apache.org/jira/browse/PODLINGNAMESEARCH-90 >>>>>>>>> <https://issues.apache.org/jira/browse/PODLINGNAMESEARCH-90>, and >>> fill >>>>>>>>> out >>>>>>>>> column C in the spreadsheet. >>>>>>>>> >>>>>>>>> Julian >>>>>>>>> >>>>>>>>> [1] https://github.com/stephenh/joist >>>>>>>>> <https://github.com/stephenh/joist> >>>>>>>>> >>>>>>>>> >>>>>>>>> On Nov 30, 2015, at 7:01 PM, Jacques Nadeau <[email protected] >>>>>>>>> <mailto:[email protected]>> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> +1 >>>>>>>>> >>>>>>>>> -- >>>>>>>>> Jacques Nadeau >>>>>>>>> CTO and Co-Founder, Dremio >>>>>>>>> >>>>>>>>> On Mon, Nov 30, 2015 at 6:34 PM, Wes McKinney <[email protected] >>>>>>>>> <mailto:[email protected]>> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> Should we have a last call for votes, closing EOD tomorrow >>> (Tuesday)? >>>>>>>>> I >>>>>>>>> missed this for a few days last week with holiday travel. >>>>>>>>> >>>>>>>>> On Thu, Nov 26, 2015 at 3:04 PM, Julian Hyde < >>> [email protected] <mailto:[email protected]>> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> Consulting a lawyer is part of the Apache branding process but the >>>>>>>>> first >>>>>>>>> stage is to gather a list of potential conflicts - >>>>>>>>> https://issues.apache.org/jira/browse/PODLINGNAMESEARCH-90 >>>>>>>>> <https://issues.apache.org/jira/browse/PODLINGNAMESEARCH-90> is an >>>>>>>>> example. >>>>>>>>> >>>>>>>>> The other part, frankly, is to pick your battles. >>>>>>>>> >>>>>>>>> A year or so ago Actian re-branded Vectorwise as Vector. >>>>>>>>> >>>>>>>>> >>>>>>>>> >>> http://www.zdnet.com/article/actian-consolidates-its-analytics-portfolio/ >>> <http://www.zdnet.com/article/actian-consolidates-its-analytics-portfolio/>. >>>>>>>>> Given that it is an analytic database in the Hadoop space I think >>>>>>>>> that is >>>>>>>>> as close to a “direct hit” as it gets. I don’t think we need a >>> lawyer >>>>>>>>> to >>>>>>>>> tell us that. Certainly it makes sense to look for conflicts for >>> the >>>>>>>>> other >>>>>>>>> alternatives before consulting lawyers. >>>>>>>>> >>>>>>>>> Julian >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> On Nov 25, 2015, at 9:42 PM, Marcel Kornacker <[email protected] >>>>>>>>> <mailto:[email protected]> >>>> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> On Tue, Nov 24, 2015 at 3:25 PM, Jacques Nadeau < >>> [email protected] <mailto:[email protected]>> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> Ok guys, >>>>>>>>> >>>>>>>>> I don't think anyone is doing a thorough analysis of viaability. I >>>>>>>>> did a >>>>>>>>> quick glance and the top one (Vector) seems like it would have an >>>>>>>>> issue >>>>>>>>> with conflict of an Actian product. The may be fine. Let's do a >>>>>>>>> second >>>>>>>>> phase vote. >>>>>>>>> >>>>>>>>> >>>>>>>>> I'm assuming you mean Vectorwise? >>>>>>>>> >>>>>>>>> Before we do anything else, could we have a lawyer look into this? >>>>>>>>> Last >>>>>>>>> time around that I remember (Parquet), Twitter's lawyers did a good >>>>>>>>> job >>>>>>>>> of >>>>>>>>> weeding out the potential trademark violations. >>>>>>>>> >>>>>>>>> Alex, could Twitter get involved this time around as well? >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> Pick your top 3 (1,2,3 with 3 being top preference) >>>>>>>>> >>>>>>>>> Let's get this done by Friday and then we can do a podling name >>>>>>>>> search >>>>>>>>> starting with the top one. >>>>>>>>> >>>>>>>>> Link again: >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>> https://docs.google.com/spreadsheets/d/1q6UqluW6SLuMKRwW2TBGBzHfYLlXYm37eKJlIxWQGQM/edit#gid=304381532&vpid=A1 >>> >>> <https://docs.google.com/spreadsheets/d/1q6UqluW6SLuMKRwW2TBGBzHfYLlXYm37eKJlIxWQGQM/edit#gid=304381532&vpid=A1> >>>>>>>>> >>>>>>>>> thanks >>>>>>>>> >>>>>>>>> >>>>>>>>> -- >>>>>>>>> Jacques Nadeau >>>>>>>>> CTO and Co-Founder, Dremio >>>>>>>>> >>>>>>>>> On Fri, Nov 20, 2015 at 9:24 AM, Jacques Nadeau < >>> [email protected] <mailto:[email protected]>> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> Ok, it looks like we have a candidate list (we actually got 11 >>> since >>>>>>>>> there was a three-way tie for ninth place): >>>>>>>>> >>>>>>>>> VectorArrowhoneycombHerringbonejoistV2Pietcolbufbatonimpulsevictor >>>>>>>>> Next we need to do trademark searches on each of these to see >>> whether >>>>>>>>> we're likely to have success. I've moved candidates to a second >>> tab: >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>> https://docs.google.com/spreadsheets/d/1q6UqluW6SLuMKRwW2TBGBzHfYLlXYm37eKJlIxWQGQM/edit#gid=304381532 >>> >>> <https://docs.google.com/spreadsheets/d/1q6UqluW6SLuMKRwW2TBGBzHfYLlXYm37eKJlIxWQGQM/edit#gid=304381532> >>>>>>>>> >>>>>>>>> Anybody want to give a hand in analyzing potential conflicts? >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> -- >>>>>>>>> Jacques Nadeau >>>>>>>>> CTO and Co-Founder, Dremio >>>>>>>>> >>>>>>>>> On Mon, Nov 16, 2015 at 12:10 PM, Jacques Nadeau < >>> [email protected] <mailto:[email protected]>> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> Everybody should pick their ten favorites using the numbers 1 to >>> 10. >>>>>>>>> >>>>>>>>> 10 is most preferred >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> -- >>>>>>>>> Jacques Nadeau >>>>>>>>> CTO and Co-Founder, Dremio >>>>>>>>> >>>>>>>>> On Mon, Nov 16, 2015 at 10:17 AM, Ted Dunning < >>> [email protected] <mailto:[email protected]>> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> >>>>>>>>> Single vote for most preferred? >>>>>>>>> >>>>>>>>> Single transferable vote? >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> On Tue, Nov 17, 2015 at 2:50 AM, Jacques Nadeau < >>> [email protected] <mailto:[email protected]>> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> Given that a bunch of people added names to the sheet, I'll take >>>>>>>>> that as tacit agreement to the proposed process. >>>>>>>>> >>>>>>>>> Let's move to the first vote phase. I've added a column for >>>>>>>>> everybody's votes. Let's try to wrap up the vote by 10am on >>>>>>>>> Wednesday. >>>>>>>>> >>>>>>>>> thanks! >>>>>>>>> Jacques >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> -- >>>>>>>>> Jacques Nadeau >>>>>>>>> CTO and Co-Founder, Dremio >>>>>>>>> >>>>>>>>> On Thu, Nov 12, 2015 at 12:03 PM, Jacques Nadeau < >>> [email protected] <mailto:[email protected]> >>>>>>>>> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> >>>>>>>>> Hey Guys, >>>>>>>>> >>>>>>>>> It sounds like we need to do a little more work on the Vector >>>>>>>>> proposal >>>>>>>>> before the board would like to consider it. The main point of >>>>>>>>> contention >>>>>>>>> right now is the name of the project. We need to decide on a name >>>>>>>>> and get >>>>>>>>> it signed off through PODLINGNAMESEARCH. >>>>>>>>> >>>>>>>>> Naming is extremely subjective so I'd like to propose a process for >>>>>>>>> selection that minimizes pain. This is an initial proposal and >>>>>>>>> >>>>>>>>> We do the naming in the following steps >>>>>>>>> - 1: Collect a set of names to be considered >>>>>>>>> - 2: Run a vote for 2 days where each member ranks their top 10 >>>>>>>>> options >>>>>>>>> 1..10 >>>>>>>>> - 3: Take the top ten vote getters and do a basic analysis of >>>>>>>>> whether we >>>>>>>>> think that any have legal issues. Keep dropping names that have >>>>>>>>> this until >>>>>>>>> we get with 10 reasonably solid candidate names >>>>>>>>> - 5: Take the top ten names and give people 48 hours to rank their >>>>>>>>> top 3 >>>>>>>>> names >>>>>>>>> - 6: Start a PODLINGNAMESEARCH on the top rank one, if that doesn't >>>>>>>>> work, >>>>>>>>> try the second and third options. >>>>>>>>> >>>>>>>>> I suggest we take name suggestions for step 1 from everyone but >>> then >>>>>>>>> constrain the voting to the newly proposed project [1]. We could >>>>>>>>> just do >>>>>>>>> this in a private email thread but I think doing it on Drill dev is >>>>>>>>> better >>>>>>>>> in the interest of transparency. This isn't the perfect place for >>>>>>>>> that but >>>>>>>>> I'm not sure a better place exists. >>>>>>>>> >>>>>>>>> I'm up for changing any or all of this depending on what others >>>>>>>>> think. Just >>>>>>>>> wanted to get the ball rolling on a proposed process. >>>>>>>>> >>>>>>>>> If this works, I've posted a doc at [2] that we can use for step 1. >>>>>>>>> >>>>>>>>> Thanks, >>>>>>>>> Jacques >>>>>>>>> >>>>>>>>> [1] List of proposed new project members/voters: Todd Lipcon, Ted >>>>>>>>> Dunning, >>>>>>>>> Michael Stack, P. Taylor Goetz, Julian Hyde, Julien Le Dem, Jacques >>>>>>>>> Nadeau, >>>>>>>>> James Taylor, Jake Luciani, Parth Chandra, Alex Levenson, Marcel >>>>>>>>> Kornacker, >>>>>>>>> Steven Phillips, Hanifi Gunes, Wes McKinney, Jason Altekruse, David >>>>>>>>> Alves, >>>>>>>>> Zain Asgar, Ippokratis Pandis, Abdel Hakim Deneche, Reynold Xin. >>>>>>>>> [2] >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>> https://docs.google.com/spreadsheets/d/1q6UqluW6SLuMKRwW2TBGBzHfYLlXYm37eKJlIxWQGQM/edit#gid=0 >>> >>> <https://docs.google.com/spreadsheets/d/1q6UqluW6SLuMKRwW2TBGBzHfYLlXYm37eKJlIxWQGQM/edit#gid=0> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> -- >>>>>>>> Alex Levenson >>>>>>>> @THISWILLWORK >>>>>>> >>>>>>> >>>> >>>> >>> >> >> >> >> -- >> Julien >> >> >
