chemistry-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Peter Monks <pmo...@alfresco.com>
Subject Re: [DISCUSSION] OpenCMIS 0.10.0
Date Mon, 12 Aug 2013 14:53:07 GMT
G'day Florian,

Yeah *I* understand that OpenCMIS is production grade - I've explicitly chosen to use it in
several Alfresco products that I manage.  The problem is when I deliver that message to other
prospective implementers it sometimes falls on deaf ears.

Having the explanation below, or Dieter's quote, or similar may help such implementers to
decide in favour of OpenCMIS.  Having a v1.0 would be more effective.

Cheers,
Peter



 
 


On Aug 12, 2013, at 4:57 AM, Florian Müller <fmui@apache.org> wrote:

> Hi Peter,
> 
> OpenCMIS consists of multiple more or less independent parts. The "serious flaw" was
that the client library couldn't connect to a Web Service endpoint anymore. That doesn't touch
the server framework or the InMemory repository or the common parser classes or anything else.
> Unfortunately, releasing OpenCMIS 0.9.1 as a bug fix release didn't work out. But that
shouldn't stop us from improving other areas.
> 
> OpenCMIS is used in many open source and commercial products and productive scenarios
today. But the JavaDoc could be improved and some code areas could need some more comments
and clean up to make it better maintainable. I think 'high quality' is also defined by these
things. Till now we have focused on making it feature complete (-> CMIS 1.1) and correct.
> Personally, I would like to address the documentation and maintainability areas before
we release v1.0, even if it doesn't change any APIs and we could theoretically do this after
v1.0 is released. But that's only my opinion. It should be a community decision.
> 
> Apart from that, I guess OpenCMIS development will not stop for a long time. At least
the TCK will grow. But I can also envision support for more authentication methods (for example
OAuth) and specific adaptations for certain environments (application servers, enterprise
service buses, JAX-WS implementations, etc.).
> 
> 
> - Florian
> 
> 
>> If v0.9 had "serious flaws", I might ask why 0.10.0 adds "a new
>> TypeFactory class and a couple of utility classes" and makes "changes
>> for cleanup spread over hundreds of classes"?  Wouldn't a more
>> conservative, fix-centric approach be more advisable?
>> 
>> Regardless, I think the comment that "I do not see any reason to
>> rush." concerns me the most.  CMIS v1.0 was released more than 3 years
>> ago and the argument has been made that there still isn't a stable,
>> reliable client library available.  Clearly Apache Chemistry is not an
>> official CMIS client library, but the goals of CMIS are hindered by
>> this lack.
>> 
>> Perhaps I'm misunderstanding the primary goal of the OpenCMIS
>> sub-project though - is it to provide high quality Java CMIS client
>> libraries, or is it more around client library experimentation?
>> 
>> Cheers,
>> Peter
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> On Aug 9, 2013, at 8:45 AM, Jay Brown <jay.brown@us.ibm.com> wrote:
>> 
>>> I agree with Jens.  Make it 0.10.0.  Getting really close though.
>>> 
>>> I will be doing a fair share of testing (server side OpenCMIS) between now and
November that once completed will give me more confidence as well.
>>> 
>>> 
>>> 
>>> 
>>> 
>>> 
>>> Jay Brown
>>> Senior Engineer, ECM Development
>>> IBM Software Group
>>> jay.brown@us.ibm.com
>>> 
>>> "Huebel, Jens" ---08/08/2013 11:37:55 PM---Personally I feel that this is 0.10
and not a 1.0 release. It appeared that the previous one 0.9 had
>>> 
>>> 
>>> From:
>>> 
>>> "Huebel, Jens" <j.huebel@sap.com>
>>> 
>>> To:
>>> 
>>> "dev@chemistry.apache.org" <dev@chemistry.apache.org>,
>>> 
>>> Date:
>>> 
>>> 08/08/2013 11:37 PM
>>> 
>>> Subject:
>>> 
>>> Re: [DISCUSSION] OpenCMIS 0.10.0
>>> 
>>> 
>>> 
>>> Personally I feel that this is 0.10 and not a 1.0 release.
>>> 
>>> It appeared that the previous one 0.9 had some serious flaws which made us
>>> releasing another version pretty soon.
>>> 
>>> With the current release we introduced a new TypeFactory class and a
>>> couple of utility classes being essential for the core functionality of a
>>> server if they are in use. This code saw the daylight only a couple of
>>> days ago and definitely needs a proof that it is reliable and stable.
>>> There also have been changes for cleanup spread over hundreds of classes.
>>> The InMemory server is not for production use and therefore is of minor
>>> importance but needs cleanup in some areas. I also feel that the
>>> documentation is not in a 1.0 state yet.
>>> 
>>> Nothing would be worse for our project than releasing a crippled 1.0
>>> release after years of effort. And Peter I fear your users will hesitate
>>> to use this stuff forever if we run 1.0 into the weeds ;)
>>> 
>>> I do not see any reason to rush. I agree to target a 1.0 release for the
>>> fall, end-of-year time frame if we do not introduce new functionality
>>> since our last 0.x version. Isn't this good style for any project?
>>> 
>>> Just my 2 cents
>>> 
>>> Jens
>>> 
>>> 
>>> On 08.08.13 20:23, "Florian Müller" <fmui@apache.org> wrote:
>>> 
>>> >We can actually do both in parallel. Our release manager can cut a
>>> >0.10.0 release. Once we have a release candidate we can work full steam
>>> >on 1.0. We don't have to wait for the release process to finish.
>>> >
>>> >Btw. Any help with the JavaDoc is very welcome. It would be great if
>>> >some native speakers could support us here.
>>> >
>>> >
>>> >- Florian
>>> >
>>> >
>>> >> Thanks Florian.  If a v1.0 is that close, I'd vote for doing whatever's
>>> >>necessary to get it to that point, even if it delays the bug fixes etc.
>>> >>a bit.  I've had a little pushback (not much, but not zero either) from
>>> >>potential users of the library because of a perception that it's
>>> >>"pre-release" (based solely on the version number, as best I can tell).
>>> >>
>>> >> Cheers,
>>> >> Peter
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>
>>> >>
>>> >> On Aug 8, 2013, at 8:32 AM, Florian Müller <fmui@apache.org>
wrote:
>>> >>
>>> >>> We have full CMIS 1.1 support now.
>>> >>> If the community feels comfortable calling it 1.0 we can do that.
Any
>>> >>>opinions?
>>> >>>
>>> >>> I think we should improve the JavaDoc to a point that it sufficiently
>>> >>>covers all public APIs and then call it 1.0. There are also some
places
>>> >>>that need cleaning. But I don't expect that we add any major
>>> >>>functionality in the near future.
>>> >>>
>>> >>>
>>> >>> - Florian
>>> >>>
>>> >>>
>>> >>>> On Thu, 8 Aug 2013, Peter Monks wrote:
>>> >>>>> +1, but as a side note, what's the gating factor on a v1.0?
>>> >>>> Full CMIS v1.1 support might seem a good reason for the version
bump?
>>> >>>>
>>> >>>> Nick
>>> >>
>>> >>
>>> 
>>> 
>>> 
> 


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message