[
https://issues.apache.org/jira/browse/HADOOP-14217?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15943031#comment-15943031
]
Steve Loughran commented on HADOOP-14217:
-----------------------------------------
Nicholas has pointed out the history of this —looks like it's old and fairly
fundamental.
Quick case-by-case workarounds aren't the right place to do it. As an example,
those globber fixes are at best going to handle the case where the last element
in a path contains a ":", code in which a parent path contains a colon, e,g
"s3a://bucket/elt1:name/child2 is probably doomed, if not immediately,
certainly when later things happen like it gets sent over the wire,
deserialized, saved to the list of files to distcp, reloaded, etc. etc.
Fixing this is a major piece of work, which will include tests (if HDFS adds
":" support, miniHDFS can be the test FS, otherwise some mock ramfs is needed).
Along with the changes to the FS spec, we'd need a contract test for all
supporting filesystems to implement, including the object stores, testing all
those corner case and operations we can think of. Probably time to do a decent
set of globber tests there too, which the contract lacks.
I think that could be good, HADOOP-3257 would the be place, and HDFS the team
to collaborate with.
> Object Storage: support colon in object path
> --------------------------------------------
>
> Key: HADOOP-14217
> URL: https://issues.apache.org/jira/browse/HADOOP-14217
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs, fs/oss
> Reporter: Genmao Yu
>
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
