accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Christopher Tubbs (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-2401) client code cannot create an RFile without logging a warning
Date Tue, 11 Mar 2014 21:53:47 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-2401?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13931009#comment-13931009
] 

Christopher Tubbs commented on ACCUMULO-2401:
---------------------------------------------

Sorry, that should have been "shell config command", not "documentation".

Experimental properties were hidden from the output of the shell config command, unless they
were actually used, by excluding them from the DefaultConfiguration. The reason is that the
semantics of "experimental" imply "optional", and "optional" implies that it could be null/unset/absent,
and still work correctly. While I'm primarily concerned about prominent exposure of experimental
properties through the shell and documentation, the same argument applies to anything that
uses the DefaultConfiguration, because that's where the semantics of "experimental" are communicated
to the user (by their omission/lack of prominence).

Because the DefaultConfiguration conveys the semantics about which properties are prominently
exposed as available, I think experimental properties should continue to be omitted from it,
and we add a test to ensure that the defaults for these are null and excluded. The code that
depends on them being otherwise should be changed to handle them being null, or they should
not be considered experimental. Alternatively, we'd have to find a better way to convey that
certain properties are experimental, in spite of being so prominently exposed (and I can't
think of a way to programatically communicate that, other than through omission; I think documentation
that says "don't use this unless you know what you're doing; this is experimental and subject
to change" is insufficient, as it will largely be ignored... if it's there, people will use
it and depend on it).

The code change, for reference:
{code}git log -n1 -p -U5 f2de427f -- core/src/main/java/org/apache/accumulo/core/conf/DefaultConfiguration.java{code}

I made ACCUMULO-2460, for follow-on.

> client code cannot create an RFile without logging a warning
> ------------------------------------------------------------
>
>                 Key: ACCUMULO-2401
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-2401
>             Project: Accumulo
>          Issue Type: Bug
>          Components: client
>    Affects Versions: 1.5.0
>            Reporter: Eric Newton
>            Assignee: Eric Newton
>              Labels: 16_qa_bug
>             Fix For: 1.5.2, 1.6.0
>
>
> Examining the output of the SimpleProxyIT, I noticed the test was always printing a message
about not being able to access accumulo-site.xml.  This is coming from a call to AccumuloConfiguration.getSiteConfiguration()
buried down in BCFile.Writer for the encryption stuff.
> Client-side code should not be calling getSiteConfiguration().



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message