[
https://issues.apache.org/jira/browse/HADOOP-1779?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12523672
]
dhruba borthakur commented on HADOOP-1779:
------------------------------------------
I think this API could potentially complicate things when we go to a more
distributed namenode. If, in future, we consider a hash function to map a
filename to the location of its namenode, then permission checking along the
entire path while looking up a file cannot be implemented efficiently.
In short, i think we need some more discussion on whether HDFS will do access
checks along the entire path while looking up a file. I believe NFS does not do
access checks along the entire path (especially because it uses fileids for
lookups).
> Small INodeDirectory enhancement to get all existing INodes components on a
> path
> --------------------------------------------------------------------------------
>
> Key: HADOOP-1779
> URL: https://issues.apache.org/jira/browse/HADOOP-1779
> Project: Hadoop
> Issue Type: Improvement
> Components: dfs
> Reporter: Christophe Taton
> Assignee: Christophe Taton
> Priority: Trivial
> Fix For: 0.15.0
>
> Attachments: 1779_2007-08-27c.patch
>
>
> This patch introduces {{INode[] INodeDirectory.getExistingPathINodes(String
> path)}}, modeled over the existing {{getNode()}} / {{getINode()}}.
> The purpose of this is to provide a way to retrieve all existing INodes in a
> path in a single tree scan.
> This to allow fast INode's permission (see
> [HADOOP-1298|https://issues.apache.org/jira/browse/HADOOP-1298]) checking
> along the path, and without requiring the {{INode.parent}} field which could
> be removed (see
> [HADOOP-1687|https://issues.apache.org/jira/browse/HADOOP-1687]).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
