Let's remove the HDFS proxy contrib component. It has not had a
meaningful contribution in almost a year.   We had to disable the
tests in HDFS-1666 because it is broken an no one has been willing to
maintain it.  Nigel started this vote previously, with the exception
of one vote everyone was +1. I tried several to follow up on the
original thread a couple of times but I never got a response so I
figured I'd start a new vote.

Btw, Alejandro is contributing a full HDFS proxy replacement (also
supports read/write, kerberos, etc) in HDFS-2178.

I'm +1.

Thanks,
Eli

1. http://mail-archives.apache.org/mod_mbox/hadoop-general/201102.mbox/%[EMAIL PROTECTED]%3E

On Jul 20, 2011, at 2:26 PM, Eli Collins wrote:
>
> Btw, Alejandro is contributing a full HDFS proxy replacement (also
> supports read/write, kerberos, etc) in HDFS-2178.

Which is basically blocked by Alfredo's lack of security review.

On Jul 20, 2011, at 2:42 PM, Allen Wittenauer wrote:

>
> On Jul 20, 2011, at 2:26 PM, Eli Collins wrote:
>>
>> Btw, Alejandro is contributing a full HDFS proxy replacement (also
>> supports read/write, kerberos, etc) in HDFS-2178.
>
> Which is basically blocked by Alfredo's lack of security review.

I'm also not comfortable with a critical part of the security infrastructure coming from a github project. Will Cloudera contribute Alfredo to Hadoop?

-- Owen

On Wed, Jul 20, 2011 at 2:45 PM, Owen O'Malley <[EMAIL PROTECTED]> wrote:
>
> On Jul 20, 2011, at 2:42 PM, Allen Wittenauer wrote:
>
>>
>> On Jul 20, 2011, at 2:26 PM, Eli Collins wrote:
>>>
>>> Btw, Alejandro is contributing a full HDFS proxy replacement (also
>>> supports read/write, kerberos, etc) in HDFS-2178.
>>
>>       Which is basically blocked by Alfredo's lack of security review.
>
> I'm also not comfortable with a critical part of the security infrastructure coming from a github project. Will Cloudera contribute Alfredo to Hadoop?
>

Yup, per my mail to Allen, Alejandro is planning to contribute Alfredo
directly to Hadoop once he's finished mavenizing common.

Going forward we need to figure out a better place to host these
components (eg Alfredo, or Hadoop libraries used by HBase) than Hadoop
because other projects that depend on these components don't need to
wait for a Hadoop release to get a new version.

Thanks,
Eli

On Jul 20, 2011, at 2:49 PM, Eli Collins wrote:

> On Wed, Jul 20, 2011 at 2:45 PM, Owen O'Malley <[EMAIL PROTECTED]> wrote:
>>
>> On Jul 20, 2011, at 2:42 PM, Allen Wittenauer wrote:
>>
>>>
>>> On Jul 20, 2011, at 2:26 PM, Eli Collins wrote:
>>>>
>>>> Btw, Alejandro is contributing a full HDFS proxy replacement (also
>>>> supports read/write, kerberos, etc) in HDFS-2178.
>>>
>>>       Which is basically blocked by Alfredo's lack of security review.
>>
>> I'm also not comfortable with a critical part of the security infrastructure coming from a github project. Will Cloudera contribute Alfredo to Hadoop?
>>
>
> Yup, per my mail to Allen, Alejandro is planning to contribute Alfredo
> directly to Hadoop once he's finished mavenizing common.

Let's get the ordering right and wait to remove HDFSProxy until Hoops and Alfredo are in.

-- Owen

On Wed, Jul 20, 2011 at 2:50 PM, Owen O'Malley <[EMAIL PROTECTED]> wrote:
>
> On Jul 20, 2011, at 2:49 PM, Eli Collins wrote:
>
>> On Wed, Jul 20, 2011 at 2:45 PM, Owen O'Malley <[EMAIL PROTECTED]> wrote:
>>>
>>> On Jul 20, 2011, at 2:42 PM, Allen Wittenauer wrote:
>>>
>>>>
>>>> On Jul 20, 2011, at 2:26 PM, Eli Collins wrote:
>>>>>
>>>>> Btw, Alejandro is contributing a full HDFS proxy replacement (also
>>>>> supports read/write, kerberos, etc) in HDFS-2178.
>>>>
>>>>       Which is basically blocked by Alfredo's lack of security review.
>>>
>>> I'm also not comfortable with a critical part of the security infrastructure coming from a github project. Will Cloudera contribute Alfredo to Hadoop?
>>>
>>
>> Yup, per my mail to Allen, Alejandro is planning to contribute Alfredo
>> directly to Hadoop once he's finished mavenizing common.
>
> Let's get the ordering right and wait to remove HDFSProxy until Hoops and Alfredo are in.
>

I don't think we should remove HDFSProxy because Hoop exists, I think
we should remove it because it's broken and not being maintained.
Broken and un-maintained code should be removed no?

Thanks,
Eli

On Jul 20, 2011, at 2:55 PM, Eli Collins wrote:
>
> I don't think we should remove HDFSProxy because Hoop exists, I think
> we should remove it because it's broken and not being maintained.
> Broken and un-maintained code should be removed no?

That'd remove large chunks of hadoop.

*rimshot*

On Wed, Jul 20, 2011 at 3:03 PM, Allen Wittenauer <[EMAIL PROTECTED]> wrote:
>
> On Jul 20, 2011, at 2:55 PM, Eli Collins wrote:
>>
>> I don't think we should remove HDFSProxy because Hoop exists, I think
>> we should remove it because it's broken and not being maintained.
>> Broken and un-maintained code should be removed no?
>
>        That'd remove large chunks of hadoop.
>

We've made good progress removing broken / unmaintained code from
Hadoop.  Common no longer contains any contrib projects. And HDFS is
down to two (hdfs proxy and fuse-dfs, and the latter is actually
maintained, by myself and others).

Thanks,
Eli

This code has been broken for a long time and, while it's still in use
in some companies, they've had a long time to offer fixes.

+1 on abandoning it.

On Wed, Jul 20, 2011 at 3:12 PM, Eli Collins <[EMAIL PROTECTED]> wrote:
> On Wed, Jul 20, 2011 at 3:03 PM, Allen Wittenauer <[EMAIL PROTECTED]> wrote:
>>
>> On Jul 20, 2011, at 2:55 PM, Eli Collins wrote:
>>>
>>> I don't think we should remove HDFSProxy because Hoop exists, I think
>>> we should remove it because it's broken and not being maintained.
>>> Broken and un-maintained code should be removed no?
>>
>>        That'd remove large chunks of hadoop.
>>
>
> We've made good progress removing broken / unmaintained code from
> Hadoop.  Common no longer contains any contrib projects. And HDFS is
> down to two (hdfs proxy and fuse-dfs, and the latter is actually
> maintained, by myself and others).
>
> Thanks,
> Eli
>

+1 let's get rid of it for Pete's sake

On Wed, Jul 20, 2011 at 03:16PM, Jakob Homan wrote:
> This code has been broken for a long time and, while it's still in use
> in some companies, they've had a long time to offer fixes.
>
> +1 on abandoning it.
>
> On Wed, Jul 20, 2011 at 3:12 PM, Eli Collins <[EMAIL PROTECTED]> wrote:
> > On Wed, Jul 20, 2011 at 3:03 PM, Allen Wittenauer <[EMAIL PROTECTED]> wrote:
> >>
> >> On Jul 20, 2011, at 2:55 PM, Eli Collins wrote:
> >>>
> >>> I don't think we should remove HDFSProxy because Hoop exists, I think
> >>> we should remove it because it's broken and not being maintained.
> >>> Broken and un-maintained code should be removed no?
> >>
> >> ═ ═ ═ ═That'd remove large chunks of hadoop.
> >>
> >
> > We've made good progress removing broken / unmaintained code from
> > Hadoop. ═Common no longer contains any contrib projects. And HDFS is
> > down to two (hdfs proxy and fuse-dfs, and the latter is actually
> > maintained, by myself and others).
> >
> > Thanks,
> > Eli
> >

+1 (for a second time).

On Jul 20, 2011, at 3:16 PM, Jakob Homan <[EMAIL PROTECTED]> wrote:

> This code has been broken for a long time and, while it's still in use
> in some companies, they've had a long time to offer fixes.
>
> +1 on abandoning it.
>
> On Wed, Jul 20, 2011 at 3:12 PM, Eli Collins <[EMAIL PROTECTED]> wrote:
>> On Wed, Jul 20, 2011 at 3:03 PM, Allen Wittenauer <[EMAIL PROTECTED]> wrote:
>>>
>>> On Jul 20, 2011, at 2:55 PM, Eli Collins wrote:
>>>>
>>>> I don't think we should remove HDFSProxy because Hoop exists, I think
>>>> we should remove it because it's broken and not being maintained.
>>>> Broken and un-maintained code should be removed no?
>>>
>>>        That'd remove large chunks of hadoop.
>>>
>>
>> We've made good progress removing broken / unmaintained code from
>> Hadoop.  Common no longer contains any contrib projects. And HDFS is
>> down to two (hdfs proxy and fuse-dfs, and the latter is actually
>> maintained, by myself and others).
>>
>> Thanks,
>> Eli
>>

I already have a patch of Alfredo with all package names and
copyright/license headers as per Apache guidelines.

Once HADOOP-6671 is in it will be straight forward to bring into Hadoop
codebase.

Thanks.

Alejandro

On Wed, Jul 20, 2011 at 2:49 PM, Eli Collins <[EMAIL PROTECTED]> wrote:

> On Wed, Jul 20, 2011 at 2:45 PM, Owen O'Malley <[EMAIL PROTECTED]>
> wrote:
> >
> > On Jul 20, 2011, at 2:42 PM, Allen Wittenauer wrote:
> >
> >>
> >> On Jul 20, 2011, at 2:26 PM, Eli Collins wrote:
> >>>
> >>> Btw, Alejandro is contributing a full HDFS proxy replacement (also
> >>> supports read/write, kerberos, etc) in HDFS-2178.
> >>
> >>       Which is basically blocked by Alfredo's lack of security review.
> >
> > I'm also not comfortable with a critical part of the security
> infrastructure coming from a github project. Will Cloudera contribute
> Alfredo to Hadoop?
> >
>
> Yup, per my mail to Allen, Alejandro is planning to contribute Alfredo
> directly to Hadoop once he's finished mavenizing common.
>
> Going forward we need to figure out a better place to host these
> components (eg Alfredo, or Hadoop libraries used by HBase) than Hadoop
> because other projects that depend on these components don't need to
> wait for a Hadoop release to get a new version.
>
> Thanks,
> Eli
>

On Wed, Jul 20, 2011 at 2:42 PM, Allen Wittenauer <[EMAIL PROTECTED]> wrote:
>
> On Jul 20, 2011, at 2:26 PM, Eli Collins wrote:
>>
>> Btw, Alejandro is contributing a full HDFS proxy replacement (also
>> supports read/write, kerberos, etc) in HDFS-2178.
>
>        Which is basically blocked by Alfredo's lack of security review.
>

My understanding is that, post common mavenization (which is almost
done), Alejandro is planning to contribute Alfredo directly to Hadoop,
where it will go through the regular review process.  Hoop is useful
w/o Alfredo though, it doesn't need to block it.

Thanks,
Eli

+1 (non binding) I am always in favor of removing dead code.  If anyone ever does want to look at it or fix it and bring it back in they can go back to an older version. That is part of what version control is for.

--
Bobby Evans

On 7/20/11 4:26 PM, "Eli Collins" <[EMAIL PROTECTED]> wrote:

Let's remove the HDFS proxy contrib component. It has not had a
meaningful contribution in almost a year.   We had to disable the
tests in HDFS-1666 because it is broken an no one has been willing to
maintain it.  Nigel started this vote previously, with the exception
of one vote everyone was +1. I tried several to follow up on the
original thread a couple of times but I never got a response so I
figured I'd start a new vote.

Btw, Alejandro is contributing a full HDFS proxy replacement (also
supports read/write, kerberos, etc) in HDFS-2178.

I'm +1.

Thanks,
Eli

1. http://mail-archives.apache.org/mod_mbox/hadoop-general/201102.mbox/%[EMAIL PROTECTED]%3E

+1 removing hdfs proxy
I do not think that Alfredo or any other project should be a precondition
for removing unused code.
--Konstantin

On Wed, Jul 20, 2011 at 2:26 PM, Eli Collins <[EMAIL PROTECTED]> wrote:

> Let's remove the HDFS proxy contrib component. It has not had a
> meaningful contribution in almost a year.   We had to disable the
> tests in HDFS-1666 because it is broken an no one has been willing to
> maintain it.  Nigel started this vote previously, with the exception
> of one vote everyone was +1. I tried several to follow up on the
> original thread a couple of times but I never got a response so I
> figured I'd start a new vote.
>
> Btw, Alejandro is contributing a full HDFS proxy replacement (also
> supports read/write, kerberos, etc) in HDFS-2178.
>
> I'm +1.
>
> Thanks,
> Eli
>
> 1.
> http://mail-archives.apache.org/mod_mbox/hadoop-general/201102.mbox/%[EMAIL PROTECTED]%3E
>

With 6 +1s (5 binding) and no -1s the vote passes.

Thanks,
Eli

On Wed, Jul 20, 2011 at 2:26 PM, Eli Collins <[EMAIL PROTECTED]> wrote:
> Let's remove the HDFS proxy contrib component. It has not had a
> meaningful contribution in almost a year.   We had to disable the
> tests in HDFS-1666 because it is broken an no one has been willing to
> maintain it.  Nigel started this vote previously, with the exception
> of one vote everyone was +1. I tried several to follow up on the
> original thread a couple of times but I never got a response so I
> figured I'd start a new vote.
>
> Btw, Alejandro is contributing a full HDFS proxy replacement (also
> supports read/write, kerberos, etc) in HDFS-2178.
>
> I'm +1.
>
> Thanks,
> Eli
>
> 1. http://mail-archives.apache.org/mod_mbox/hadoop-general/201102.mbox/%[EMAIL PROTECTED]%3E
>