[Building Sakai] [Deploying Sakai] MT deprecation recommendations for 2.7 (fwd)

[Ian Boston]

On 12 Mar 2010, at 17:53, Stephen Marquard wrote:

> [I didn't see the original of this on the management list - perhaps Aaron isn't subscribed to that or something and it bounced?]
> 
> JCR is used by at least one contrib tool (NYU's Memory Palace). It was included in the kernel originally so that tools could use it if they wished to. I think this needs a clearer survey of current use and intentions for future use by contrib tools before it gets removed. I'm referring here to the JCR service itself rather than the CHS JCR adapter.

The CHS JCR [1] should be disabled and removed if no one wants to support, its in "experimental" anyway, but I don't think Aaron was talking about this (although item 6 mixed the 2 parts up)

The JCR Service [2] has been running in production at Cambridge since July 2008.
NYU is also running a version of it that has been ported to Xythos.

It is *not* enabled by default [3, line 198] unless you set jcr.experimental=true, so it does not consume memory at runtime.
It does include the Jackrabbit core jars, but these are not loaded if not enabled.

I see the last commit was about 10 months ago, (Aaron you fixed a possible NPE as part of a static code review.)
From jira it has 0 bugs reported pre K1 and 1 in K1 [4] although I think that was miss-reported. (searched KNL for JCR)

------------------------

Having said all that, its only about 10 very simple classes and could easily be moved out of kernel into contrib. and I am sure no one would notice it moving, but you should do a release of it as soon as its moved.


1 https://source.sakaiproject.org/svn//kernel/trunk/kernel-impl-experimental/content/
2 https://source.sakaiproject.org/svn//kernel/trunk/kernel-impl/src/main/java/org/sakaiproject/jcr/
3 https://source.sakaiproject.org/viewsvn/kernel/trunk/kernel-impl/src/main/java/org/sakaiproject/jcr/jackrabbit/RepositoryBuilder.java?revision=60847&view=markup
4 http://jira.sakaiproject.org/browse/KNL-363
> 
> The case for removing Presentation seems the weakest. Saying "it is not maintained" isn't really an argument because presumably if it isn't maintained, it would fall into the remit of the MT to do so. So an argument that "it is not maintainable" would be stronger but I don't think that's the case here and it doesn't exactly have a lot of open JIRAs (5 major, 2 minor). So then it comes down to whether it's simply not used by enough institutions to see it retained in the release, and I'd prefer to see some actual figures on that (e.g. how many sites are using it in 2.5 / 2.6 who would be affected by a decision to drop it from 2.7).
> 
> Also I'm assuming that the MT is only making these recommendations now because of its relative newness, but for future release cycles, these decisions should really be made around the same time as the tool promotion decisions. Every tool or feature removed from the release potentially represents additional work for migrating sites to keep that functionality if they need to.
> 
> Regards
> Stephen
> 
> 

<snip>

> 6) JCR - The JCR/jackrabbit module in the kernel is a maintenance
> burden and does not appear to be used. It increases the memory profile
> and complexity of the kernel even when not in use. We recommend it be
> removed from the release. JCR was meant as a possible replacement for
> the content hosting service but this was not completed and there is no
> one to maintain this code.


<snip>