Mailing-List: contact hadoop-dev-help@lucene.apache.org; run by ezmlm
Precedence: bulk
Reply-To: hadoop-dev@lucene.apache.org
Received-SPF: neutral (herse.apache.org: local policy)
Message-ID: <461FAFEE.3040207@apache.org>
Date: Fri, 13 Apr 2007 09:29:34 -0700
From: Doug Cutting <cutting@apache.org>
User-Agent: Thunderbird 1.5.0.10 (X11/20070306)
MIME-Version: 1.0
To: hadoop-dev@lucene.apache.org
Subject: Re: [jira] Commented: (HADOOP-1214) the first step for streaming
 clean up
References: <985608.1176412455491.JavaMail.jira@brutus>
 <2f6eb616632ea9aa7935b6c1dc02afdf@yahoo-inc.com>
 <461EBE2E.6080208@apache.org>
 <c5b291899ba41b67aeb0cf9a0ca04bd7@yahoo-inc.com>
In-Reply-To: <c5b291899ba41b67aeb0cf9a0ca04bd7@yahoo-inc.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Arkady Borkovsky wrote:
>> On Apr 12, 2007, at 4:18 PM, Doug Cutting wrote:
>> The new classes in question are not a part of streaming, but are being 
>> added to the mapred package.
> Is not Hadoop Streaming part of Hadoop MapReduce product?

Streaming is currently in contrib not in the core.

I'm just suggesting we use consistent, accurate and descriptive 
terminology within the core.  These classes to not read nor generate 
lines.  They do facilitate interoperability with other line-based tools 
like TextInputFormat and TextOutputFormat.

> And are not the classes in question supposed to be referred to by "naive 
> users" on the Hadoop Streaming command line?

I don't think we should name core classes to make the streaming command 
line more intuitive.  If all else were equal, sure, that's a good thing, 
but, core classes should be named as consistently, accurately and 
descriptively as possible.  If streaming's command line is confusing, 
then that should be fixed in streaming, no?

Doug