[HADOOP-1677] improve semantics of the hadoop dfs command

• Type: Improvement
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 0.17.0
• Assignee: Mahadev Konar
• Reporter: Nigel Daley
• Created: 03/Aug/07 21:35
• Updated: 02/May/13 06:29
• Resolved: 21/Feb/08 01:48

Description

HADOOP-230 cover syntax changes. Capturing here some DFS command line semantic comments between Doug and I:

The goal for DFS should be "UNIX-like when possible". The goals of HDFS are primarily to be usable, scalable, reliable, high-performance, and, secondarily, not to be gratuitously incompatible with UNIX. So, when it's easy to be compatible, we certainly should. But when UNIX compatibility fights with one of those other goals, it may lose.

Two examples of incompatibilities with UNIX (that will need to be evaluated against the above goal), are:

UNIX: cp foo bar succeeds if bar exists
DFS: dfs -put foo bar fails if bar exists. (same for dfs -cp).

UNIX: mkdir foo fails if foo exists
DFS: dfs -mkdir succeeds silently if foo exists and is a directory (if foo is a file than it fails).

Many more such incompatibilities exist.

Issue Links

• depends upon: HADOOP-2190 dfs ls and lsr commands differ from POSIX standards
• depends upon: HADOOP-2191 dfs du and dus commands differ from POSIX standards
• depends upon: HADOOP-2192 dfs mv command differs from POSIX standards
• depends upon: HADOOP-2193 dfs rm and rmr commands differ from POSIX standards
• depends upon: HADOOP-2195 dfs mkdir command differs from POSIX standards