Come back to clear out the codebase and IP concerns.
[Benedikt] I still have concerns about the IP, since this seems to be an Intel
codebase.
I have checked this. Chimera was developed as Apache 2 License from ground up.
Agree with Jochen that the license matters.
Internally, this was approved as part of larger open source efforts on Apache
Hadoop and related projects in Intel. We have IP plan considered as part the
open source process.
As to the codebase, such as the package name is com.intel prefixed, it was
technical decision when using com.intel.chimera as the package prefix. We
original planned to use org.apache.chimera prefix. But we found that we
couldnot publish org.apache. grouped artifacts to maven central repository,
which needs to somewhat ownership for org.apache domain.
To resolve the codebase problem, once all things are ready from Commons, we
rename in a branch. And the branched code can be copied into Commons github as
final.
Thanks,
Haifeng
-Original Message-
From: Chen, Haifeng [mailto:haifeng.c...@intel.com]
Sent: Wednesday, February 24, 2016 12:40 PM
To: Commons Developers List
Cc: common-...@hadoop.apache.org
Subject: RE: [crypto][chimera] Next steps
>> The same should be there with Chimera/Apache Crypto.
Yes, current implementation will fallback to JCE Cipher if native is not
available.
[Uma] we would fix up IP issues if any sooner. If you see all the code file
license header is with Apache License files.
The current repo and package structure there with name Intel. I will check with
Haifeng on resolution of this.
I will work with this ASAP for clear this out.
Thanks,
Haifeng
-Original Message-
From: Gangumalla, Uma [mailto:uma.ganguma...@intel.com]
Sent: Wednesday, February 24, 2016 5:13 AM
To: Commons Developers List
Cc: common-...@hadoop.apache.org
Subject: Re: [crypto][chimera] Next steps
Thanks all for the valuable feedbacks and discussions.
Here are my replies for some of the questions..
[Mark wrote]
It depends. I care less about the quality of the code than I do about the
community that comes with it / forms around it. A strong community can fix code
issues. Great code can't save a weak community.
[uma] Nice point. Fully agreed to it.
[Jochen wrote]
Therefore, I suggest that you provide at least fallback implementations in pure
Java, which are being used, if the JNI based stuff is not available (for
whatever reason).
[Uma] Thank you for the suggestion Jochen, If I understand your point right,
Yes its there in Hadoop when we develop.
Here is the JIRA HADOOP-10735 : Fall back AesCtrCryptoCodec implementation from
OpenSSL to JCE if non native support.
The same should be there with Chimera/Apache Crypto.
[Benedikt]
I still have concerns about the IP, since this seems to be an Intel codebase. I
do not have the necessary experience to say what would be the right way here.
My gut feeling tells me, that we should go through the incubator. WDYT?
And [Jochen wrote]
"An Intel codebase" is not a problem as such. Question is: "Available under
what license?"
[Uma] we would fix up IP issues if any sooner. If you see all the code file
license header is with Apache License files.
The current repo and package structure there with name Intel. I will check with
Haifeng on resolution of this.
[Jochen wrote]
So, have the Chimera project attempt to resolve them quickly. If
possible: Fine. If not: We still have the Incubator as a possibility.
[Uma] Agree. We would resolve on this points in sooner.
Regards,
Uma
On 2/23/16, 1:18 AM, "Mark Thomas" wrote:
>On 23/02/2016 09:12, sebb wrote:
>> On 23 February 2016 at 07:34, Benedikt Ritter
>>wrote:
>>> I'm confused. None of the other PMC members has expressed whether he
>>>or she want's the see Chimera/crypto joining Apache Commons, yet
>>>we're already discussing how JNI bindings should be handled.
>>>
>>> I'd like to see:
>>> 1) a clear statement whether Chimera/crypto should become part of
>>>Apache Commons. Do we need a vote for that?
>>
>> Yes, of course.
>>
>> However that decision clearly depends on at least some of the design
>> aspects of the code.
>> If it were written entirely in C or Fortran, it would not be a
>> suitable candidate.
>>
>>> 2) Discuss a plan on how to do that (I've described a possible plan
>>>[1])
>>> 3) After that is clear: discuss design details regarding the component.
>>
>> Some design details impact on the decision.
>>
>> Indeed even for pure Java code the code quality has a bearing on
>> whether Commons would/could want to take it.
>> Would we want a large code base with no unit-tests, no build
>> mechanism, and no comments?
>
>It depends. I care less about the quality of the code than I do about
>the community that comes with it / forms around it. A strong community
>can fix code issues. Great code can't save a weak community.
>
>How about creating a new sandbox component, let folks start work and
>see how the community develops?
>
>
