jackrabbit-oak-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "angela (JIRA)" <j...@apache.org>
Subject [jira] [Comment Edited] (OAK-8234) Reduce object allocation in PermissionProviderImpl for admin sessions
Date Mon, 15 Apr 2019 09:06:00 GMT

    [ https://issues.apache.org/jira/browse/OAK-8234?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16817666#comment-16817666
] 

angela edited comment on OAK-8234 at 4/15/19 9:05 AM:
------------------------------------------------------

[~stillalex], while the intended change makes a lot of sense to me, i don't particularly like
the delegatee approach with yet another redirection. in particular since it creates a lot
of changes to the {{PermissionProviderImpl}} that might result in subtle regressions. Is there
a reason for not making the distinction between administrative subjects and regular subjects
right in the {{AuthorizationConfigurationImpl}}? that would place that fundamental difference
between set of principals that are subjected to permission evaluation vs those that are not
prominently in the flow. also for administrative principals the whole {{Mount}} handling is
IMO redundant and which is not taken into account in the patch as far as i saw. wdyt?

regarding the failing test: please don't mark tests as ignored. if we decide that the test
is not relevant for administrative principals, we should move it to a test for non-administrative
principals... since a lot of tests for the current {{PermissionProvider}} are writting for
an admin user, we might refactor the tests a well if administrative subjects get another instance
of {{PermissionProvider}}.  the overall branch coverage of the permission package is 86% today
and we should IMO improve that before making such a major refactoring. 


was (Author: anchela):
[~stillalex], while the intended change makes a lot of sense to me, i don't particularly like
the delegatee approach with yet another redirection. in particular since it creates a lot
of changes to the {{PermissionProviderImpl}} that might result in subtle regressions. Is there
a reason for not making the distinction between administrative subjects and regular subjects
right in the {{AuthorizationConfigurationImpl}}? that would place that fundamental difference
between set of principals that are subjected to permission evaluation vs those that are not
prominently in the flow. also for administrative principals the whole {{Mount}} handling is
IMO redundant and which is not taken into account in the patch as far as i saw. wdyt?

> Reduce object allocation in PermissionProviderImpl for admin sessions
> ---------------------------------------------------------------------
>
>                 Key: OAK-8234
>                 URL: https://issues.apache.org/jira/browse/OAK-8234
>             Project: Jackrabbit Oak
>          Issue Type: Improvement
>          Components: core, security
>            Reporter: Alex Deparvu
>            Assignee: Alex Deparvu
>            Priority: Major
>
> There was already a lot of work done here to improve efficiency of admin sessions, but
we can still do better by simply not creating many of the objects needed as parameters for
the {{CompiledPermissions}} which will be ignored anyway in the case of admin sessions.
> Ex: There are a lot of immutable trees (created via {{PermissionUtil.getReadOnlyTree(tree,
immutableRoot)}}) that are not used for the evaluation.
> Thanks to [~mreutegg] for spotting this one!



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message