Well, I spoke two soon. There are two pending tickets on my plate (Accumulo-404, running on top of kerberoized hdfs and Accumulo-489, passwords are not secured with input format) that are a lot easier to implement, if not only feasible, against 20.203+ (CDH3u0+). It is possible to do version checking within the code and then calling via reflection, but that is ugly, ugly code that I think we'd be best to avoid. So I ask the community, do we have any reasons NOT to migrate our minimum version to 20.203 now that we have a need for a newer version?
John On Tue, Mar 20, 2012 at 4:29 PM, Robert Vesse <[email protected]> wrote: > I was just interested since 0.20.2 is now two years old if there was any > compelling reason (other than QA requirements) to stay with that version > > Rob > > Rob Vesse -- YarcData.com -- A Division of Cray Inc > Software Engineer, Bay Area > m: 925.960.3941 | o: 925.264.4729 | @: [email protected] | Skype: > rvesse > 6210 Stoneridge Mall Rd | Suite 120 | Pleasanton CA, 94588 > > > On Mar 20, 2012, at 12:26 PM, John Vines wrote: > > We don't depend on any specific features of 20.205, so I don't know why > we need to update the minimum version. I know many of us have been working > with 20.205 without issue. > > John > > Sent from my phone, so pardon the typos and brevity. > On Mar 20, 2012 2:50 PM, "Robert Vesse" <[email protected]> wrote: > >> I haven't looked at the 1.4 branch in detail, did you guys move to a >> newer version of Hadoop as well, the 1.3.5 release uses the comparatively >> ancient 0.20.2 version of Hadoop >> >> Rob Vesse -- YarcData.com -- A Division of Cray Inc >> Software Engineer, Bay Area >> m: 925.960.3941 | o: 925.264.4729 | @: [email protected] | Skype: >> rvesse >> 6210 Stoneridge Mall Rd | Suite 120 | Pleasanton CA, 94588 >> >> >> On Mar 19, 2012, at 5:36 PM, Eric Newton wrote: >> >> Yes, the 1.4 branch is considered stable. >> >> We've been pounding on 1.4 for quite a while. The scale of the testing >> has over a longer period of time, more complete and more aggressive. >> Another team has been using it for some benchmark testing at scale, and it >> is remarkable for the *lack* of problems. >> >> No, you cannot mix 1.3 clients with 1.4 servers. We jumped from 0.3 of >> thrift to 0.6.1, and that was a huge change for us; I'm not even sure if it >> can be on-the-wire compatible. Our goal is to ensure compatibility for >> 1.X.* level for any constant X. But a switch from 1.X to 1.Y... we'll >> remove deprecated APIs, and break on-the-wire compatibility. >> >> -Eric >> >> On Mon, Mar 19, 2012 at 1:57 PM, Robert Vesse <[email protected]>wrote: >> >>> Is the 1.4.0 branch considered stable enough for use? And is it the >>> API compatible between 1.3.5 and 1.4.0 or do both the database and the >>> client need to be on the exact same version for communication to work >>> correctly? >>> >>> Rob >>> >>> On Mar 19, 2012, at 10:06 AM, John Vines wrote: >>> >>> I believe we don't have them distributed because our poms do not include >>> adequate licensing info for us to publicly release them. 1.4.0 should be >>> released when we release it though. >>> >>> John >>> >>> On Mon, Mar 19, 2012 at 12:58 PM, Robert Vesse <[email protected]>wrote: >>> >>>> Is there a reason why Maven artifacts for 1.3.5 are not available in >>>> the Apache repositories? >>>> >>>> Which Maven repositories (if any) are they available in? >>>> >>>> Rob >>>> >>> >>> >>> >> >> >
