jackrabbit-oak-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael Dürig <mdue...@apache.org>
Subject [NEWS] Oak flash (3rd edition)
Date Mon, 01 Oct 2012 15:55:46 GMT

Hi,

Last week has been pretty busy with the Oak Hackathon on Tuesday 
followed by the .adaptTo 2012 conference (http://adaptto.mixxt.de/) 
Wednesday to Friday in Berlin.

* The Hackathon 
(http://wiki.apache.org/jackrabbit/Oak%20Hackathon%20September%202012) 
was quite a success even though not too much hacking has actually taken 
place. Half of the roughly 10 attendees where new to the project so we 
decided to turn this more into an intro session. Jukka and Michael gave 
a preview on their Oak presentation for .adaptTo 2012 
(http://goo.gl/7fikE) and Jukka went through some coding examples 
showing how to write a custom commit hook for Oak. The code examples 
will become parts of the NodeState model documentation. See 
https://github.com/apache/jackrabbit-oak/blob/trunk/doc/nodestate.md.
We briefly discussed about repeating this kind of event and I think we 
agreed that this would be a good thing. I will follow up in a separate 
thread to sort out the details regarding time frame and location.

* Jukka's and Michael's presentation "Jackrabbit Oak - the next 
generation content repository" (http://goo.gl/7fikE) generated quite 
some interest at .adaptTo 2012. Most questions revolved around issues 
related to backward compatibility, better scalability in cluster 
deployments and the impact of snapshot isolation.

* The documentation on the NodeState has further evolved. Most notably 
there is a new section on how to build new node states using the node 
state builder. See 
https://github.com/apache/jackrabbit-oak/blob/trunk/doc/nodestate.md.

(Note that the Github mirror lags a bit behind. So if the relevant 
content is not their yet use this svn link: 
http://svn.apache.org/viewvc/jackrabbit/oak/trunk/doc/nodestate.md?view=markup)

* Last week's testing resulted in a flurry of filed issues on the 
QueryEngine parsing and interpretation of results. We uncovered some 
inconsistencies on node type handling (OAK-325), full text condition 
parsing(OAK-), like on path constraints (OAK-347) and a bunch of JCR 
functions that are not supported yet in Oak (excerpt, defer, similar, eq).
As work on the query engine itself stabilises there is a bit of advance 
happening on the custom query index side: Lucene is chugging along 
nicely (OAK-340, OAK-334), and more interestingly we have a Solr 
integration up for review at OAK-307, take a look if interested!
A gentle reminder, the main JIRA issue that tracks query related 
problems and inconsistencies compared to jackrabbit is OAK-237.

* Tudor Rogoz contributed an initial version of a performance test suite 
for the Microkernel which can be used for benchmarking the various 
implementations we have. We envision to use this also for scalability 
testing of cluster deployments further down the line. See OAK-335 for 
the respective JIRA issue and 
http://svn.apache.org/viewvc?rev=1392315&view=rev for the contributions 
itself.

* With the resolution of OAK-197 there are some minor changes in the Oak 
API: the ConflictHandler argument was removed from the Root.commit() and 
Root.rebase() methods. We figured that these parameters unnecessarily 
complicate the client API and break encapsulation and thus decided to 
move the conflict handling logic to a plugin instead of exposing it this 
way through the API.

* In revision 1392352 
(http://svn.apache.org/viewvc?rev=1392352&view=rev) Marcel moved the 
repository setup code from ContentRepositoryImpl and ContentSessionImpl 
to its own plugin InitialContent. The refactored OSGi related classes 
allow to hook this directly into the life cycle of the Microkernel.

* On the debating side there was a discussion on whether and how we want 
to support virtual content in Oak 
(http://markmail.org/message/3fpssbdv6srabwx6). Not much progress so far 
but I expect this to be a topic which will pop up rather sooner again 
than later.
A further - rather lengthy debate - was on custom index configuration 
(http://markmail.org/message/bdgi77dd6wy2hkbp). As far as I could follow 
it was mainly about whether to store the index (configurations) 
centrally or locally with the content. While we currently don't have an 
implementation for the non centralised case, I think we can expect some 
experimental ones soon. This will probably allow us to make a better 
informed decision on which way(s) to go.
Also there was a discussion to go beyond mere path base access in the 
Microkernel (http://markmail.org/message/vwq22osa7q2anm33). The 
motivations for bringing this up was that path based access doesn't seem 
to scale too well in the MongoDB based implementation.
Finally a rather recent discussion is concerned about the destiny of Oak 
(http://markmail.org/message/ga4mn2x2xqsvzwg6) and is all about whether 
Oak should graduate into its own project or become Jackrabbit 3.0 
replacing the current Jackrabbit trunk. While the discussion is still 
ongoing most of the opinions raised so far seem to be in favour of the 
2nd option.

Thanks to Alex for providing the query/indexing section of this write up.

cheers,
Michael





Mime
View raw message