On Sat, Mar 21, 2015 at 2:47 AM, Branko Čibej <[email protected]> wrote:
> On 21.03.2015 05:25, Dmitriy Setrakyan wrote: > > On Fri, Mar 20, 2015 at 7:33 PM, Marvin Humphrey <[email protected] > > > > wrote: > > > >> On Fri, Mar 20, 2015 at 6:53 PM, Dmitriy Setrakyan > >> <[email protected]> wrote: > >> > >>> I think we made a mistake and imported the wrong version of the source > >>> code. The same version is provided here by Doug Lee without the GPL > >>> headers, but only with Public Domain header: > >>> http://gee.cs.oswego.edu/cgi-bin/viewcvs.cgi/jsr166/src/jsr166e/ > >>> > >>> I guess we will have to resubmit the release, unless you can accept it > in > >>> the current shape. Please advise. > >> If the files don't contain any GPL IP, then what we have is a "licensing > >> documentation bug": there's public domain IP which is perfectly fine to > >> include, but it's misleadingly labeled. Shipping a release which > contains > >> such IP does not pose any legal problems. > >> > >> Contrast that with a licensing error, such as GPL IP onto which someone > has > >> slapped an ALv2 header. Shipping such a release could leave users and > >> redistributors open to a claim of copyright violation. > >> > >> Licensing documentation bugs can vary from inconsequential to > catastrophic > >> depending on how badly they mislead downstream consumers. It seems to > me > >> that > >> while this one might cause alarm, it shouldn't cause anybody to do > anything > >> illegal. So long as you are *certain* that those *exact* versions of > the > >> files are available under Doug Lea's public domain dedication and > contain > >> no > >> GPL mods, I think it's OK. > >> > > Thanks for quick response. > > > > I am *certain* because these files were initially grabbed from Doug Lea's > > JSR 166 page. The wrong license headers were added by mistake. > > Do you happen to know exactly which versions of those files were > imported? I looked at the ConcurrentHashMap implementation and compared > the current version on that site: > > > http://gee.cs.oswego.edu/cgi-bin/viewcvs.cgi/jsr166/src/jsr166e/ConcurrentHashMapV8.java?revision=1.123 > > > with the one in the Ignite release package, and the differences are far > larger than just license header changes: > > https://paste.apache.org/p/3SAz > > The code for JSR-166 has evolved significantly since we did the initial import close to 2 years ago, before the JDK8 was officially released yet. I also tried to find the exact version we imported, but at this point I am afraid it would be almost impossible. On top of that we have removed/changed some not-needed code in the migrated classes. > > Sorry, I should've checked this much earlier ... > > I would suggest to *not* just remove the license headers in those files. > Instead, delete the current files and import a fresh set directly from > the repository at gee.cs.oswego.edu, then mention in NOTICE that they > were imported from there, not from OpenJDK. Also please add a README > file to the directory where those files are imported and note the exact > versions of the imported files (maybe best to just list the ViewCVS > download URLs, like the one above). > Yes, absolutely agree. I have created a ticket in Ignite JIRA, IGNITE-545, to address this. > > Once you've imported the original files, you can, e.g., change the > package name and make other minor modifications. The point is to have an > audit trail of changes from the public-domain original to whatever we > ship; currently, the Git log shows no such trail. > Yes, agree. Unfortunately this code was moved and renamed several times since it was initially migrated. I am pretty sure that GIT origins of the code are not clearanymore. > -- Brane > >
