You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hbase.apache.org by Todd Lipcon <to...@cloudera.com> on 2010/05/20 19:44:54 UTC
Version number of next release
Hi HBasers,
Time for the second proposal of the day!
I'd like to start a discussion around the version number of the upcoming
"durable HBase" release. The release I'm referring to is the one currently
being worked towards on trunk, and the one that FB and Cloudera plan to work
with for production clusters round about Q3 2010.
The current name for this release is 0.21. I think this is going to cause
user confusion due to the previous "lockstep versioning" that HBase has had
with regard to Hadoop. I think many people will assume they need to use
Hadoop 0.21 (being billed as an unstable release at least for 0.21.0) and
generally not quite understand why our version number is the same if we have
no tie to the Hadoop version. So, I am generally -1 on calling this next
HBase release 0.21.0.
The other factor is that I think we all see this upcoming release as a major
step up from 0.20. Namely, it provides true durability of every write, much
improved cluster stability, a new build system, replication, and countless
other improvements that everyone's been cranking on. I'm sure given the
number of people now working on the project, we'll see even a few more great
improvements pop up before we're ready to freeze.
Some have suggested we jump all the way to HBase 1.0. I think this is a bit
ambitious, as 1.0 implies a level of API stability we're not quite ready to
commit to. Perhaps we can go there some time next year, but don't want to
open that can of worms yet :)
So, beyond not liking either 0.21 or 1.0, I don't have a strong opinion.
Some have suggested 0.90, as it is lexically much bigger than 0.20 but
clearly not 1.0 yet. Others have suggested 0.30, to give us room to go to
0.40, 0.50, etc before a 1.0.
Thoughts?
-Todd
--
Todd Lipcon
Software Engineer, Cloudera
Re: Version number of next release
Posted by Jean-Daniel Cryans <jd...@apache.org>.
I mean, +1 on 0.66
On Thu, May 20, 2010 at 10:54 PM, Jean-Daniel Cryans
<jd...@apache.org> wrote:
> +1
>
> On Thu, May 20, 2010 at 10:50 PM, Stack <st...@duboce.net> wrote:
>> Lets not call it 0.21. I wanted to call it 0.66.0 so we could do a
>> logo for it: http://people.apache.org/~stack/66.jpg
>>
>> I'm good w/ 0.90.0 or 0.30.0.
>>
>> St.Ack
>>
>> On Thu, May 20, 2010 at 10:44 AM, Todd Lipcon <to...@cloudera.com> wrote:
>>> Hi HBasers,
>>>
>>> Time for the second proposal of the day!
>>>
>>> I'd like to start a discussion around the version number of the upcoming
>>> "durable HBase" release. The release I'm referring to is the one currently
>>> being worked towards on trunk, and the one that FB and Cloudera plan to work
>>> with for production clusters round about Q3 2010.
>>>
>>> The current name for this release is 0.21. I think this is going to cause
>>> user confusion due to the previous "lockstep versioning" that HBase has had
>>> with regard to Hadoop. I think many people will assume they need to use
>>> Hadoop 0.21 (being billed as an unstable release at least for 0.21.0) and
>>> generally not quite understand why our version number is the same if we have
>>> no tie to the Hadoop version. So, I am generally -1 on calling this next
>>> HBase release 0.21.0.
>>>
>>> The other factor is that I think we all see this upcoming release as a major
>>> step up from 0.20. Namely, it provides true durability of every write, much
>>> improved cluster stability, a new build system, replication, and countless
>>> other improvements that everyone's been cranking on. I'm sure given the
>>> number of people now working on the project, we'll see even a few more great
>>> improvements pop up before we're ready to freeze.
>>>
>>> Some have suggested we jump all the way to HBase 1.0. I think this is a bit
>>> ambitious, as 1.0 implies a level of API stability we're not quite ready to
>>> commit to. Perhaps we can go there some time next year, but don't want to
>>> open that can of worms yet :)
>>>
>>> So, beyond not liking either 0.21 or 1.0, I don't have a strong opinion.
>>> Some have suggested 0.90, as it is lexically much bigger than 0.20 but
>>> clearly not 1.0 yet. Others have suggested 0.30, to give us room to go to
>>> 0.40, 0.50, etc before a 1.0.
>>>
>>> Thoughts?
>>>
>>> -Todd
>>>
>>> --
>>> Todd Lipcon
>>> Software Engineer, Cloudera
>>>
>>
>
Re: Version number of next release
Posted by Jean-Daniel Cryans <jd...@apache.org>.
+1
On Thu, May 20, 2010 at 10:50 PM, Stack <st...@duboce.net> wrote:
> Lets not call it 0.21. I wanted to call it 0.66.0 so we could do a
> logo for it: http://people.apache.org/~stack/66.jpg
>
> I'm good w/ 0.90.0 or 0.30.0.
>
> St.Ack
>
> On Thu, May 20, 2010 at 10:44 AM, Todd Lipcon <to...@cloudera.com> wrote:
>> Hi HBasers,
>>
>> Time for the second proposal of the day!
>>
>> I'd like to start a discussion around the version number of the upcoming
>> "durable HBase" release. The release I'm referring to is the one currently
>> being worked towards on trunk, and the one that FB and Cloudera plan to work
>> with for production clusters round about Q3 2010.
>>
>> The current name for this release is 0.21. I think this is going to cause
>> user confusion due to the previous "lockstep versioning" that HBase has had
>> with regard to Hadoop. I think many people will assume they need to use
>> Hadoop 0.21 (being billed as an unstable release at least for 0.21.0) and
>> generally not quite understand why our version number is the same if we have
>> no tie to the Hadoop version. So, I am generally -1 on calling this next
>> HBase release 0.21.0.
>>
>> The other factor is that I think we all see this upcoming release as a major
>> step up from 0.20. Namely, it provides true durability of every write, much
>> improved cluster stability, a new build system, replication, and countless
>> other improvements that everyone's been cranking on. I'm sure given the
>> number of people now working on the project, we'll see even a few more great
>> improvements pop up before we're ready to freeze.
>>
>> Some have suggested we jump all the way to HBase 1.0. I think this is a bit
>> ambitious, as 1.0 implies a level of API stability we're not quite ready to
>> commit to. Perhaps we can go there some time next year, but don't want to
>> open that can of worms yet :)
>>
>> So, beyond not liking either 0.21 or 1.0, I don't have a strong opinion.
>> Some have suggested 0.90, as it is lexically much bigger than 0.20 but
>> clearly not 1.0 yet. Others have suggested 0.30, to give us room to go to
>> 0.40, 0.50, etc before a 1.0.
>>
>> Thoughts?
>>
>> -Todd
>>
>> --
>> Todd Lipcon
>> Software Engineer, Cloudera
>>
>
Re: Version number of next release
Posted by Stack <st...@duboce.net>.
Lets not call it 0.21. I wanted to call it 0.66.0 so we could do a
logo for it: http://people.apache.org/~stack/66.jpg
I'm good w/ 0.90.0 or 0.30.0.
St.Ack
On Thu, May 20, 2010 at 10:44 AM, Todd Lipcon <to...@cloudera.com> wrote:
> Hi HBasers,
>
> Time for the second proposal of the day!
>
> I'd like to start a discussion around the version number of the upcoming
> "durable HBase" release. The release I'm referring to is the one currently
> being worked towards on trunk, and the one that FB and Cloudera plan to work
> with for production clusters round about Q3 2010.
>
> The current name for this release is 0.21. I think this is going to cause
> user confusion due to the previous "lockstep versioning" that HBase has had
> with regard to Hadoop. I think many people will assume they need to use
> Hadoop 0.21 (being billed as an unstable release at least for 0.21.0) and
> generally not quite understand why our version number is the same if we have
> no tie to the Hadoop version. So, I am generally -1 on calling this next
> HBase release 0.21.0.
>
> The other factor is that I think we all see this upcoming release as a major
> step up from 0.20. Namely, it provides true durability of every write, much
> improved cluster stability, a new build system, replication, and countless
> other improvements that everyone's been cranking on. I'm sure given the
> number of people now working on the project, we'll see even a few more great
> improvements pop up before we're ready to freeze.
>
> Some have suggested we jump all the way to HBase 1.0. I think this is a bit
> ambitious, as 1.0 implies a level of API stability we're not quite ready to
> commit to. Perhaps we can go there some time next year, but don't want to
> open that can of worms yet :)
>
> So, beyond not liking either 0.21 or 1.0, I don't have a strong opinion.
> Some have suggested 0.90, as it is lexically much bigger than 0.20 but
> clearly not 1.0 yet. Others have suggested 0.30, to give us room to go to
> 0.40, 0.50, etc before a 1.0.
>
> Thoughts?
>
> -Todd
>
> --
> Todd Lipcon
> Software Engineer, Cloudera
>
Re: Version number of next release
Posted by Todd Lipcon <to...@cloudera.com>.
On Thu, May 20, 2010 at 11:01 AM, Jonathan Gray <jg...@facebook.com> wrote:
> Agreed on confusion regardless what we pick. But not sure I see what the
> value is in sticking to 0.21? Hadoop is also releasing a "development 0.21"
> release and people will think these two things are tied together, though
> they are not.
>
> I'm a big +1 on 0.90 for the next release. For the development release, we
> could call it 0.80 or 0.89.
>
I'm with you - +1 on 0.90 for the next release, 0.89 for the development
series. Let's start talking about 1.0 after we get 0.90 into production in
some real use cases.
-Todd
>
> I agree we're not 1.0 yet but feel strongly that we're nearing it and the
> next release is getting us most of the way there. Once we stabilize it out
> in the world and add whatever big features we think are left, I'm confident
> we can get to 1.0 soon after. We're already fairly stable from an API
> perspective and I don't see any big changes in the pipeline that would
> prevent backwards compatibility.
>
> So in moving towards 1.0, I'd like to get our version numbers up to point
> to that fact, thus my vote for 0.90.
>
> IMO 0.30 says "we're not tied to hadoop anymore" and "we're bigger than
> 0.20" but not much else. 0.90 says we're approaching a 1.0 and HBase is
> becoming a legitimate piece of software you can trust your data in. The
> wave of development happening right now and the production use cases that
> will be propping up in the second half of this year confirm this.
>
> JG
>
> > -----Original Message-----
> > From: Todd Lipcon [mailto:todd@cloudera.com]
> > Sent: Thursday, May 20, 2010 10:54 AM
> > To: dev@hbase.apache.org
> > Subject: Re: Version number of next release
> >
> > On Thu, May 20, 2010 at 10:48 AM, Andrew Purtell
> > <ap...@apache.org>wrote:
> >
> > > > The current name for this release is 0.21. I think this is going to
> > cause
> > > user confusion due to the previous "lockstep versioning" that HBase
> > has had
> > > with regard to Hadoop. I think many people will assume they need to
> > use
> > > Hadoop 0.21 (being billed as an unstable release at least for 0.21.0)
> > and
> > > generally not quite understand why our version number is the same if
> > we have
> > > no tie to the Hadoop version. So, I am generally -1 on calling this
> > next
> > > HBase release 0.21.0.
> > >
> > > I think there is going to be some level of confusion no matter what
> > and
> > > calling our next as-proposed unstable release series 0.21 makes sense
> > to me
> > > at least.
> > >
> > >
> > I agree - I should have included that in this email, but wanted to
> > divorce
> > the two discussions (naive of me to think that could happen ;-) )
> >
> > I think for the development series, we are going to be discouraging new
> > users from downloading and trying it unless they're fine working
> > through
> > some problems. So a bit of confusion on a dev release series is no big
> > deal.
> > The "big" release, though, should be clearly separated.
> >
> > -Todd
> >
> > --
> > Todd Lipcon
> > Software Engineer, Cloudera
>
--
Todd Lipcon
Software Engineer, Cloudera
RE: Version number of next release
Posted by Jonathan Gray <jg...@facebook.com>.
Agreed on confusion regardless what we pick. But not sure I see what the value is in sticking to 0.21? Hadoop is also releasing a "development 0.21" release and people will think these two things are tied together, though they are not.
I'm a big +1 on 0.90 for the next release. For the development release, we could call it 0.80 or 0.89.
I agree we're not 1.0 yet but feel strongly that we're nearing it and the next release is getting us most of the way there. Once we stabilize it out in the world and add whatever big features we think are left, I'm confident we can get to 1.0 soon after. We're already fairly stable from an API perspective and I don't see any big changes in the pipeline that would prevent backwards compatibility.
So in moving towards 1.0, I'd like to get our version numbers up to point to that fact, thus my vote for 0.90.
IMO 0.30 says "we're not tied to hadoop anymore" and "we're bigger than 0.20" but not much else. 0.90 says we're approaching a 1.0 and HBase is becoming a legitimate piece of software you can trust your data in. The wave of development happening right now and the production use cases that will be propping up in the second half of this year confirm this.
JG
> -----Original Message-----
> From: Todd Lipcon [mailto:todd@cloudera.com]
> Sent: Thursday, May 20, 2010 10:54 AM
> To: dev@hbase.apache.org
> Subject: Re: Version number of next release
>
> On Thu, May 20, 2010 at 10:48 AM, Andrew Purtell
> <ap...@apache.org>wrote:
>
> > > The current name for this release is 0.21. I think this is going to
> cause
> > user confusion due to the previous "lockstep versioning" that HBase
> has had
> > with regard to Hadoop. I think many people will assume they need to
> use
> > Hadoop 0.21 (being billed as an unstable release at least for 0.21.0)
> and
> > generally not quite understand why our version number is the same if
> we have
> > no tie to the Hadoop version. So, I am generally -1 on calling this
> next
> > HBase release 0.21.0.
> >
> > I think there is going to be some level of confusion no matter what
> and
> > calling our next as-proposed unstable release series 0.21 makes sense
> to me
> > at least.
> >
> >
> I agree - I should have included that in this email, but wanted to
> divorce
> the two discussions (naive of me to think that could happen ;-) )
>
> I think for the development series, we are going to be discouraging new
> users from downloading and trying it unless they're fine working
> through
> some problems. So a bit of confusion on a dev release series is no big
> deal.
> The "big" release, though, should be clearly separated.
>
> -Todd
>
> --
> Todd Lipcon
> Software Engineer, Cloudera
Re: Version number of next release
Posted by Todd Lipcon <to...@cloudera.com>.
On Thu, May 20, 2010 at 10:48 AM, Andrew Purtell <ap...@apache.org>wrote:
> > The current name for this release is 0.21. I think this is going to cause
> user confusion due to the previous "lockstep versioning" that HBase has had
> with regard to Hadoop. I think many people will assume they need to use
> Hadoop 0.21 (being billed as an unstable release at least for 0.21.0) and
> generally not quite understand why our version number is the same if we have
> no tie to the Hadoop version. So, I am generally -1 on calling this next
> HBase release 0.21.0.
>
> I think there is going to be some level of confusion no matter what and
> calling our next as-proposed unstable release series 0.21 makes sense to me
> at least.
>
>
I agree - I should have included that in this email, but wanted to divorce
the two discussions (naive of me to think that could happen ;-) )
I think for the development series, we are going to be discouraging new
users from downloading and trying it unless they're fine working through
some problems. So a bit of confusion on a dev release series is no big deal.
The "big" release, though, should be clearly separated.
-Todd
--
Todd Lipcon
Software Engineer, Cloudera
Re: Version number of next release
Posted by Andrew Purtell <ap...@apache.org>.
> The current name for this release is 0.21. I think this is going to cause user confusion due to the previous "lockstep versioning" that HBase has had with regard to Hadoop. I think many people will assume they need to use Hadoop 0.21 (being billed as an unstable release at least for 0.21.0) and generally not quite understand why our version number is the same if we have no tie to the Hadoop version. So, I am generally -1 on calling this next HBase release 0.21.0.
I think there is going to be some level of confusion no matter what and calling our next as-proposed unstable release series 0.21 makes sense to me at least.
- Andy