[
https://issues.apache.org/jira/browse/HADOOP-4044?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12637663#action_12637663
]
Doug Cutting commented on HADOOP-4044:
--------------------------------------
> so.. this is a performance compromise. But the arguments till now have been
> about "perfect" ...
This issue has long been performance-related. The assumption is that it is
unacceptable for the addition of symlinks to double or more the number of RPCs
for common HDFS operations when no links are present. I have noted this
assumption on several occasions and no one has disputed it. Do you agree with
this assumption, or do you feel we should examine it more?
> in the patch it is not clalled 'nextLinkOrOpenImpl()'.. it is just called
> 'openImpl()'.
I thought the point of contention is not the name of the method or return value
but whether to use an exception or a data structure to pass the return value.
(That pseudo code also referred to FileHandle, which is also not in the patch.)
Would you like to propose better names for the methods and objects in the
patch? I am not wedded to the names used in the patch, but perhaps we should
resolve the exception-related issue first?
> This is obviously an exaggerated example but hopefully makes a point.
I'm not sure what the point is. If the RPC system wished to return some
generic data for all protocols then it could not force a different return type,
since return types are protocol-specific, so we'd probably need to add a method
to the RPC runtime. If the situation is unusual, then an exception would be
more appropriate.
HTTP is perhaps an analogy. It uses the return code to indicate a redirect.
URLConnection.html#getInputStream() doesn't throw an exception when it
encounters a redirect. It either follows the redirect or returns the text of
the redirect page, depending on whether you've set
HttpURLConnection#setFollowRedirects. Redirects are part of the protocol. The
HTTP return type is complex. That's the model we must move towards if we wish
to handle links without adding lots more RPCs.
> Create symbolic links in HDFS
> -----------------------------
>
> Key: HADOOP-4044
> URL: https://issues.apache.org/jira/browse/HADOOP-4044
> Project: Hadoop Core
> Issue Type: New Feature
> Components: dfs
> Reporter: dhruba borthakur
> Assignee: dhruba borthakur
> Attachments: symLink1.patch, symLink1.patch, symLink4.patch,
> symLink5.patch, symLink6.patch, symLink8.patch, symLink9.patch
>
>
> HDFS should support symbolic links. A symbolic link is a special type of file
> that contains a reference to another file or directory in the form of an
> absolute or relative path and that affects pathname resolution. Programs
> which read or write to files named by a symbolic link will behave as if
> operating directly on the target file. However, archiving utilities can
> handle symbolic links specially and manipulate them directly.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
