[
https://issues.apache.org/jira/browse/OAK-6516?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16114539#comment-16114539
]
Thomas Mueller commented on OAK-6516:
-------------------------------------
> treat the index-update like a post-commit-hook that is only executed after
> all validation has been performed
That sounds good to me. I don't see how indexing could affect for example
security.
The only open question I have is possible impact on commit performance,
specially large commits (large transactions) that don't fit in the cache, and
commits with many child nodes. I guess (not sure) the main aspect is how many
times Oak goes through all operations (number of diffs). Maybe we need
additional tests for this. I'm not saying changing the order is _bad_ (I have
no idea, maybe it is even good). But there is a risk of making it worse I
guess. But maybe this is just because it's not fully clear to me how things
work exactly _right now_...
> Consider reordering the index-update-hook to reside with other
> post-commit-hooks
> --------------------------------------------------------------------------------
>
> Key: OAK-6516
> URL: https://issues.apache.org/jira/browse/OAK-6516
> Project: Jackrabbit Oak
> Issue Type: Task
> Components: core, indexing
> Reporter: angela
> Assignee: angela
>
> Currently the index-update hook is located together with other commit
> validators and is followed by the security related validators that are
> plugged to the final commit hook upon each call to {{Root.commit}} to allow
> for proper pluggability of security modules.
> Consequently a given commit may fail with some security related validation
> only after the index updated has been proceeded already.
> I would like to discuss the possibility to treat the index-update like a
> post-commit-hook that is only executed after all validation has been
> performed. In the security area this is achieved by collecting the validator
> separately from the post-commit-hooks.
> [~stillalex], [~tmueller], I would appreciate if we could discuss this topic
> as well as possible implications/regressions that might be associated with it.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
