+1 from this non-committer to holding the cache work for now. A simple on-off 
switch seems absolutely requisite, and preferably we would have fine-grained 
controls. The sudden appearance of an always-on cache like this would represent 
really surprising new behavior for integrators. Also, from the discussion on 
the PR in question [1] it seems to me that there are still some questions of 
design.

[1] https://github.com/apache/jena/pull/95

---
A. Soroka
The University of Virginia Library

> On Apr 29, 2016, at 12:38 PM, Andy Seaborne <[email protected]> wrote:
> 
> I'd like to go ahead with 3.1.0.
> 
> The work on a query cache for Fuseki [*] was tested by Osma and found to be 
> useful. There was also feedback on configuration and bounding the resources 
> consumed by the caches.  Without the ability to configure it (including turn 
> it off) I don't think that putting it in right before a release is a good 
> idea.  There are is still some discussion to be had about the detailed design 
> goals.
> 
> So what I suggest is release Jena 3.1.0 now as-is, and start a 
> testing/evaluation phase for a configurable cache.
> 
> As a cache is most important under load, and Fuseki is used in deployed and 
> public facing systems, I'm concerned that we block people picking up all the 
> other things in 3.1.0.  I think it would be good to get wider testing via 
> users@.  That will take time. (And then if people don't test it until after 
> its in a release ... well, we can only do our best.)
> 
>       Andy
> 
> [*]
> https://issues.apache.org/jira/browse/JENA-626
> https://github.com/apache/jena/pull/95
> 
> On 18/04/16 12:35, Osma Suominen wrote:
>> Hi Andy!
>> 
>> I'm happy with the current state of Jena and would support a release.
>> This time I've been more diligent in using recent snapshots so there
>> should be less chance of problems that surface only around release time.
>> 
>> I have no outstanding work that I want to do for jena-text currently.
>> JENA-1134/AnalyzingQueryParser (which you can consider including in the
>> release notes) was the major item for me since the previous release,
>> plus some bugfixes that were done earlier.
>> 
>> -Osma
>> 
>> On 18/04/16 13:44, Andy Seaborne wrote:
>>> Version 3.1.0 has quite a few good things in it1
>>> 
>>> I think it is about time we released it.
>>> 
>>> So this is a first ping for anything you want to get done for it.
>>> 
>>> As for timing - I can RM it sometime in the next few weeks though it
>>> will have to fit around other things.
>>> 
>>>     Andy
>>> 
>>> -------------------------------------------
>>> 
>>> * New custom functions and aggregate functions
>>>   Added:
>>>   + afn:sprintf (contribution from Alessandro Seganti)
>>>   + The XQuery/XPath Functions and Operators "math:" functions
>>>   + Custom aggregates for stdev etc. (also STDEV etc as keywords).
>>> 
>>> * RC features in 3.0.1:
>>>   + In-memory transactional dataset (Adam Soroka )
>>>   + SPARQL extension for CONSTRUCT Quads (Qihong Lin)
>>> 
>>> * OSGi fixes (Jaroslav Pullmann)
>>> 
>>> * Upgrades:
>>>   jsonld-java : 2.8.2
>>>     jackson 2.6.3
>>>   slf4j 1.7.20
>>>   dexx collections 0.2 -> 0.6 (OSGi)
>>> 
>>> * DatasetGraphs & transactions
>>> 
>>> * New module jena-cmds
>>> 
>>> * Logging - log4j marked <optional>
>>> 
>>> * Fuseki: Multiple service per file, shared datasets
>>> 
>>> * JSON result fix type: "literal" not "type": "typed-literal"
>>> 
>>> * Space saving when parsing (FactoryRDF)
>>>   Parsing RDF now saves space by partial interning RDFTerms
>>>   created during a each parser run.
>> 
>> 
> 

Reply via email to