You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@couchdb.apache.org by Chris Anderson <jc...@gmail.com> on 2009/01/05 03:39:28 UTC
change planning
On Sun, Jan 4, 2009 at 1:10 PM, Christopher Lenz <cm...@gmx.de> wrote:
>
> I'd like to throw in a bit of caution here.
I'm still trying to get a handle on what the appropriate procedures
ought to be. The way these last two API changes happened is I found
myself with some time, and remembered that both changes are ones that
Damien has been calling for for some time, so I double checked with
him, and then implemented.
It's hard to find the balance between getting things done quickly, and
making sure that every change is discussed openly on the ML as its
happening. My impression was the both the _slow_view and count =>
limit changes had already been discussed and decided on. But if Chris
Lenz feels differently, then we should try to avoid repeating this.
There are times when one of the committers has the opportunity and
inclination to bust through a bunch of changes. It'd be nice not to
slow the actual coding work down to discuss every patch, and as this
example shows us, it's hard to predict which patches will be
uncontroversial.
I don't think doing work like this in branches is the best idea -
where that leads to down the road is trunk lagging behind a
development branch while we discuss which changes to "backport".
I think some of this is a symptom of the fact that at least Jan, Noah,
Damien, and I are on IRC pretty much every day, so it's easy for us to
think we've got consensus on something without bringing it up on the
lists. Perhaps the best solution is to push this activity back to the
lists, with perhaps of list of decisions that have been made (even if
not implemented) maintained either in the bug tracker or on the wiki,
with reference to ML threads.
However, all that sounds very formal and slow, and at this point in a
project with only 5 committers and very simple code base and feature
set, which is not yet 1.0, I'm inclined to think erring on the side of
faster/looser may be better.
Especially some input from Apache veterans would be very helpful!
Chris
--
Chris Anderson
http://jchris.mfdz.com
Re: change planning
Posted by Noah Slater <ns...@apache.org>.
On Mon, Jan 05, 2009 at 07:25:45AM -0500, Geir Magnusson Jr. wrote:
> I'm not saying that IRC isn't very useful. it is. But when important
> technical issues are discussed on IRC, it limits who can really
> participate and collaborate in the project.
Well, this will teach me to respond to emails in reverse chronological order.
Sure, as I clarified in my previous email to the breaking changes thread, I
think that any proposed changes that break backwards compatibility or implement
major design decisions should be discussed on the mailing list before being done.
>> A blanket dismissal of IRC as a counter-productive medium, which is where
>> this whole discussion seems to be going, is prescriptive nonsense.
>
> No one is doing that.
>
> I'm saying it's important to keep the major discussions on technical
> issues on the dev list for lots of good reasons - transparency,
> participation, archival reasons.... etc
Sure, I was just worried about the thread heading towards the conclusion that no
important discussion should take place on IRC at all, as I have seen some people
allude to in the past. I would like to see any important IRC discussions "bubble
up" to the mailing list for proper consensus building.
--
Noah Slater, http://tumbolia.org/nslater
Re: change planning
Posted by "Geir Magnusson Jr." <ge...@pobox.com>.
On Jan 5, 2009, at 7:15 AM, Noah Slater wrote:
>> On Jan 5, 2009, at 4:13 AM, Geir Magnusson Jr. wrote:
>> IMO, IRC is a horrible medium for anything but direct support of
>> users, and
>> small debates, because it tends to leave people out of the
>> discussion and
>> decision-making process.
>
> On Mon, Jan 05, 2009 at 06:55:05AM -0500, Robert Dionne wrote:
>> I'll second that. IRC seems to encourage more out of the box and less
>> deliberate thinking.
>
> I am going to have to disagree with both of these statements.
>
> You, personally, might not find IRC very useful, but I do, and I
> know that some
> of the other community members do -- not just for support and
> general chat, but
> as an incredibly fluid medium for discussion and collaboration. I
> appreciate
> that mailing lists come with a host of benefits, but they do so at
> the cost of
> formality and latency.
I'm not saying that IRC isn't very useful. it is. But when
important technical issues are discussed on IRC, it limits who can
really participate and collaborate in the project.
>
>
> A blanket dismissal of IRC as a counter-productive medium, which is
> where this
> whole discussion seems to be going, is prescriptive nonsense.
No one is doing that.
I'm saying it's important to keep the major discussions on technical
issues on the dev list for lots of good reasons - transparency,
participation, archival reasons.... etc
geir
Re: change planning
Posted by Robert Dionne <bo...@gmail.com>.
point taken. I certainly didn't mean to dismiss it entirely and I do
find it enormously useful, especially as a newbie. I was just
thinking along the lines of "ready, aim, fire" and the need to do
those in the correct order, etc...
On Jan 5, 2009, at 7:15 AM, Noah Slater wrote:
>> On Jan 5, 2009, at 4:13 AM, Geir Magnusson Jr. wrote:
>> IMO, IRC is a horrible medium for anything but direct support of
>> users, and
>> small debates, because it tends to leave people out of the
>> discussion and
>> decision-making process.
>
> On Mon, Jan 05, 2009 at 06:55:05AM -0500, Robert Dionne wrote:
>> I'll second that. IRC seems to encourage more out of the box and less
>> deliberate thinking.
>
> I am going to have to disagree with both of these statements.
>
> You, personally, might not find IRC very useful, but I do, and I
> know that some
> of the other community members do -- not just for support and
> general chat, but
> as an incredibly fluid medium for discussion and collaboration. I
> appreciate
> that mailing lists come with a host of benefits, but they do so at
> the cost of
> formality and latency.
>
> A blanket dismissal of IRC as a counter-productive medium, which is
> where this
> whole discussion seems to be going, is prescriptive nonsense.
>
> --
> Noah Slater, http://tumbolia.org/nslater
Re: change planning
Posted by Jan Lehnardt <ja...@apache.org>.
Hi Anthony,
On 5 Jan 2009, at 13:35, Antony Blakey wrote:
> Couch is an outstanding reinterpretation of Notes, and I'm full of
> admiration for what's been achieved, Along with the kudos of being
> an Apache project however, comes a need to open up the clique, and
> adopt a different style of operation.
Can we offer you a place in the clique to change your opinion? :)
--
Just kidding of course. I don't think there is a "clique" here. What
happened with the renaming often happened before and that is
nothing we should be worried about. It is not that we released a
version of CouchDB with a undecided-on API.
We can still discuss the _temp_view change (+1 for `rm`-ing them
completely). Chris even brought it up on this list prior to the commit
(admittedly with a bit of a sneaky timing around new years :).
--
Thank you Anthony for putting so much time into helping us improving
CouchDB and the community!
Cheers
Jan
--
Re: change planning
Posted by Noah Slater <ns...@apache.org>.
On Tue, Jan 06, 2009 at 12:11:31AM +1030, Antony Blakey wrote:
> Couch had gone from Damien, to Damien + a small group. That group is now the
> PMC, and I think you've become used to/optimized your working in a certain
> way, with communication strategies and habits that work for that context, and
> rely on your history of interaction and collegial familiarity.
Which is great!
> Your comment about how convenient you find IRC reflects that I think.
I find IRC to be an extremely rewarding medium for me to communicate through,
and this extends well beyond the CouchDB community.
> That's what I mean by clique. It's in reference to this issue, but I'm also
> aware of the possible disruption and discomfort that transitioning to a larger
> democratic framework can entail.
There's no discomfort, and there's no real change in the organisation of the
project contributers. Within the context of this discussion, the only pertinent
change that joining the ASF as forced is the transparency of the PMC, which I
think we all agree is good thing. I think the original change without discussion
on the mailing list was a small mistake in that regard, but we're all new to
this so the whole thing is a learning experience for us.
> I see a parallel with venture capital - *my* baby was no longer my own,
> and the ideals that I wanted my company to embody were not neccessarily
> the guiding force. Maybe I'm projecting, but I'd be feeling that.
Not really, I don't sense a feeling of ownership over CouchDB. I also don't
sense any real problem with the Apache Way, just the occasional confusion or
mistake adjusting to a slightly more formal way of organising ourselves. We all
knew what we were getting into before we joined!
--
Noah Slater, http://tumbolia.org/nslater
Re: change planning
Posted by Antony Blakey <an...@gmail.com>.
On 05/01/2009, at 11:27 PM, Noah Slater wrote:
> Your framing of the issue puzzles me though. I get the impression
> that you think
> the PMC is a clique, in the general case, and that this is a wider
> issue that
> needs to be addressed. Sorry if I am reading too much into your mail.
Couch had gone from Damien, to Damien + a small group. That group is
now the PMC, and I think you've become used to/optimized your working
in a certain way, with communication strategies and habits that work
for that context, and rely on your history of interaction and
collegial familiarity. Your comment about how convenient you find IRC
reflects that I think. That's what I mean by clique. It's in reference
to this issue, but I'm also aware of the possible disruption and
discomfort that transitioning to a larger democratic framework can
entail.
I see a parallel with venture capital - *my* baby was no longer my
own, and the ideals that I wanted my company to embody were not
neccessarily the guiding force. Maybe I'm projecting, but I'd be
feeling that.
Antony Blakey
--------------------------
CTO, Linkuistics Pty Ltd
Ph: 0438 840 787
Reflecting on W.H. Auden's contemplation of 'necessary murders' in the
Spanish Civil War, George Orwell wrote that such amorality was only
really possible, 'if you are the kind of person who is always
somewhere else when the trigger is pulled'.
-- John Birmingham, "Appeasing Jakarta"
Re: change planning
Posted by Noah Slater <ns...@apache.org>.
On Mon, Jan 05, 2009 at 11:05:09PM +1030, Antony Blakey wrote:
> The Apache Way is to have decision making using mailing lists. They
> discourage IRC as a medium in which decisions are made. Mailing lists
> are good for archiving discussion history, are a more democratic and
> less 'if you happen to be watching while the discussion happens' forum,
> and allow for extended discussions over time with high visibility.
>
> Consider the issue that motivated this thread. Chris remembered that
> Damien had said he would like _temp_view to become _slow_view. He had a
> moment so he checked with Damien and suddenly it's done. Fast and Loose -
> yes. A number of people think that change wasn't/isn't a good idea, and
> would have liked a chance to argue that point. That's good for CouchDB.
> Now, as the PMC you are technically able to dismiss everyone else's view
> and simply act as Damien's support team (and it has been suggested by a
> PMC member that you operate on that basis), but that's not the Apache
> Way, and that doesn't sound much like an Apache project.
>
> Couch is an outstanding reinterpretation of Notes, and I'm full of
> admiration for what's been achieved, Along with the kudos of being an
> Apache project however, comes a need to open up the clique, and adopt a
> different style of operation.
I agree and that think this was a mistake we need to learn from.
Your framing of the issue puzzles me though. I get the impression that you think
the PMC is a clique, in the general case, and that this is a wider issue that
needs to be addressed. Sorry if I am reading too much into your mail.
--
Noah Slater, http://tumbolia.org/nslater
Re: change planning
Posted by Antony Blakey <an...@gmail.com>.
On 05/01/2009, at 10:45 PM, Noah Slater wrote:
>> On Jan 5, 2009, at 4:13 AM, Geir Magnusson Jr. wrote:
>> IMO, IRC is a horrible medium for anything but direct support of
>> users, and
>> small debates, because it tends to leave people out of the
>> discussion and
>> decision-making process.
>
> On Mon, Jan 05, 2009 at 06:55:05AM -0500, Robert Dionne wrote:
>> I'll second that. IRC seems to encourage more out of the box and less
>> deliberate thinking.
>
> I am going to have to disagree with both of these statements.
>
> You, personally, might not find IRC very useful, but I do, and I
> know that some
> of the other community members do -- not just for support and
> general chat, but
> as an incredibly fluid medium for discussion and collaboration. I
> appreciate
> that mailing lists come with a host of benefits, but they do so at
> the cost of
> formality and latency.
>
> A blanket dismissal of IRC as a counter-productive medium, which is
> where this
> whole discussion seems to be going, is prescriptive nonsense.
The Apache Way is to have decision making using mailing lists. They
discourage IRC as a medium in which decisions are made. Mailing lists
are good for archiving discussion history, are a more democratic and
less 'if you happen to be watching while the discussion happens'
forum, and allow for extended discussions over time with high
visibility.
Consider the issue that motivated this thread. Chris remembered that
Damien had said he would like _temp_view to become _slow_view. He had
a moment so he checked with Damien and suddenly it's done. Fast and
Loose - yes. A number of people think that change wasn't/isn't a good
idea, and would have liked a chance to argue that point. That's good
for CouchDB. Now, as the PMC you are technically able to dismiss
everyone else's view and simply act as Damien's support team (and it
has been suggested by a PMC member that you operate on that basis),
but that's not the Apache Way, and that doesn't sound much like an
Apache project.
Couch is an outstanding reinterpretation of Notes, and I'm full of
admiration for what's been achieved, Along with the kudos of being an
Apache project however, comes a need to open up the clique, and adopt
a different style of operation.
Antony Blakey
-------------
CTO, Linkuistics Pty Ltd
Ph: 0438 840 787
Did you hear about the Buddhist who refused Novocain during a root
canal?
His goal: transcend dental medication.
Re: change planning
Posted by Noah Slater <ns...@apache.org>.
> On Jan 5, 2009, at 4:13 AM, Geir Magnusson Jr. wrote:
> IMO, IRC is a horrible medium for anything but direct support of users, and
> small debates, because it tends to leave people out of the discussion and
> decision-making process.
On Mon, Jan 05, 2009 at 06:55:05AM -0500, Robert Dionne wrote:
> I'll second that. IRC seems to encourage more out of the box and less
> deliberate thinking.
I am going to have to disagree with both of these statements.
You, personally, might not find IRC very useful, but I do, and I know that some
of the other community members do -- not just for support and general chat, but
as an incredibly fluid medium for discussion and collaboration. I appreciate
that mailing lists come with a host of benefits, but they do so at the cost of
formality and latency.
A blanket dismissal of IRC as a counter-productive medium, which is where this
whole discussion seems to be going, is prescriptive nonsense.
--
Noah Slater, http://tumbolia.org/nslater
Re: change planning
Posted by Robert Dionne <bo...@gmail.com>.
I'll second that. IRC seems to encourage more out of the box and less
deliberate thinking.
I'll also agree with Chris that branches are not desirable. They seem
to work great on Github for experimental work and fleshing out of
ideas but in the main project resources, I think, are too limited to
support separate lines of work.
How about discussion on the mailing list followed by a summary posted
to a WIKI page immediately prior to development to clarify what folks
are voting for or against,
Bob
On Jan 5, 2009, at 4:13 AM, Geir Magnusson Jr. wrote:
> IMO, IRC is a horrible medium for anything but direct support of
> users, and small debates, because it tends to leave people out of
> the discussion and decision-making process.
>
> I don't think that it's easy to review IRC logs, because they are
> more conversational and "of the moment" rather than having even the
> minimal structure of an email discussion. it's also hard to add to
> the discussion later, unlike email...
>
> geir
>
> On Jan 5, 2009, at 4:05 AM, James Arthur wrote:
>
>> One approach might be to summarise (or 'shaila') IRC discussions
>> and circulate on the mailing list. A combination of IRC logs and
>> shailas can help discussions jump across time zones.
>>
>> Takes some work from a willing volunteer though (and thus is
>> vulnerable to stopping happening).
>>
>> Antony Blakey wrote:
>>>
>>>
>>> On 05/01/2009, at 1:09 PM, Chris Anderson wrote:
>>>
>>>> I think some of this is a symptom of the fact that at least Jan,
>>>> Noah,
>>>> Damien, and I are on IRC pretty much every day, so it's easy for
>>>> us to
>>>> think we've got consensus on something without bringing it up on
>>>> the
>>>> lists. Perhaps the best solution is to push this activity back
>>>> to the
>>>> lists, with perhaps of list of decisions that have been made
>>>> (even if
>>>> not implemented) maintained either in the bug tracker or on the
>>>> wiki,
>>>> with reference to ML threads.
>>>>
>>>> However, all that sounds very formal and slow, and at this point
>>>> in a
>>>> project with only 5 committers and very simple code base and
>>>> feature
>>>> set, which is not yet 1.0, I'm inclined to think erring on the
>>>> side of
>>>> faster/looser may be better.
>>>
>>> I think pre-1.0 is the very time that a more transparent
>>> operation should be encouraged, because non-backwards compatible
>>> change after 1.0 is considerably more difficult. And given the
>>> decree that 0.9 is the compatibility freeze point, that seems
>>> imminent.
>>>
>>> Synchronous discussion such as IRC encourages a single-time-zone
>>> of interaction.
>>>
>>> Antony Blakey
>>> -------------
>>> CTO, Linkuistics Pty Ltd
>>> Ph: 0438 840 787
>>>
>>> One should respect public opinion insofar as is necessary to
>>> avoid starvation and keep out of prison, but anything that goes
>>> beyond this is voluntary submission to an unnecessary tyranny.
>>> -- Bertrand Russell
>>>
>>>
>>>
>>
>> --
>> James Arthur
>> Founder
>> <mime-attachment.png>
>> 16 Badswell Lane
>> Appleton
>> Oxford OX13 5JN
>> UK
>>
>> Tel: +44 (0)1865 862 382
>>
>> http://www.straptank.com
>>
>
Re: change planning
Posted by "Geir Magnusson Jr." <ge...@pobox.com>.
IMO, IRC is a horrible medium for anything but direct support of
users, and small debates, because it tends to leave people out of the
discussion and decision-making process.
I don't think that it's easy to review IRC logs, because they are more
conversational and "of the moment" rather than having even the minimal
structure of an email discussion. it's also hard to add to the
discussion later, unlike email...
geir
On Jan 5, 2009, at 4:05 AM, James Arthur wrote:
> One approach might be to summarise (or 'shaila') IRC discussions and
> circulate on the mailing list. A combination of IRC logs and
> shailas can help discussions jump across time zones.
>
> Takes some work from a willing volunteer though (and thus is
> vulnerable to stopping happening).
>
> Antony Blakey wrote:
>>
>>
>> On 05/01/2009, at 1:09 PM, Chris Anderson wrote:
>>
>>> I think some of this is a symptom of the fact that at least Jan,
>>> Noah,
>>> Damien, and I are on IRC pretty much every day, so it's easy for
>>> us to
>>> think we've got consensus on something without bringing it up on the
>>> lists. Perhaps the best solution is to push this activity back to
>>> the
>>> lists, with perhaps of list of decisions that have been made (even
>>> if
>>> not implemented) maintained either in the bug tracker or on the
>>> wiki,
>>> with reference to ML threads.
>>>
>>> However, all that sounds very formal and slow, and at this point
>>> in a
>>> project with only 5 committers and very simple code base and feature
>>> set, which is not yet 1.0, I'm inclined to think erring on the
>>> side of
>>> faster/looser may be better.
>>
>> I think pre-1.0 is the very time that a more transparent operation
>> should be encouraged, because non-backwards compatible change after
>> 1.0 is considerably more difficult. And given the decree that 0.9
>> is the compatibility freeze point, that seems imminent.
>>
>> Synchronous discussion such as IRC encourages a single-time-zone of
>> interaction.
>>
>> Antony Blakey
>> -------------
>> CTO, Linkuistics Pty Ltd
>> Ph: 0438 840 787
>>
>> One should respect public opinion insofar as is necessary to avoid
>> starvation and keep out of prison, but anything that goes beyond
>> this is voluntary submission to an unnecessary tyranny.
>> -- Bertrand Russell
>>
>>
>>
>
> --
> James Arthur
> Founder
> <mime-attachment.png>
> 16 Badswell Lane
> Appleton
> Oxford OX13 5JN
> UK
>
> Tel: +44 (0)1865 862 382
>
> http://www.straptank.com
>
Re: change planning
Posted by James Arthur <th...@straptank.com>.
One approach might be to summarise (or 'shaila
<http://inamidst.com/stuff/esp/shaila>') IRC discussions and circulate
on the mailing list. A combination of IRC logs and shailas can help
discussions jump across time zones.
Takes some work from a willing volunteer though (and thus is vulnerable
to stopping happening).
Antony Blakey wrote:
>
> On 05/01/2009, at 1:09 PM, Chris Anderson wrote:
>
>> I think some of this is a symptom of the fact that at least Jan, Noah,
>> Damien, and I are on IRC pretty much every day, so it's easy for us to
>> think we've got consensus on something without bringing it up on the
>> lists. Perhaps the best solution is to push this activity back to the
>> lists, with perhaps of list of decisions that have been made (even if
>> not implemented) maintained either in the bug tracker or on the wiki,
>> with reference to ML threads.
>>
>> However, all that sounds very formal and slow, and at this point in a
>> project with only 5 committers and very simple code base and feature
>> set, which is not yet 1.0, I'm inclined to think erring on the side of
>> faster/looser may be better.
>
> I think pre-1.0 is the very time that a more transparent operation
> should be encouraged, because non-backwards compatible change after
> 1.0 is considerably more difficult. And given the decree that 0.9 is
> the compatibility freeze point, that seems imminent.
>
> Synchronous discussion such as IRC encourages a single-time-zone of
> interaction.
>
> Antony Blakey
> -------------
> CTO, Linkuistics Pty Ltd
> Ph: 0438 840 787
>
> One should respect public opinion insofar as is necessary to avoid
> starvation and keep out of prison, but anything that goes beyond this
> is voluntary submission to an unnecessary tyranny.
> -- Bertrand Russell
>
>
>
--
James Arthur
Founder
The Strap Tank <http://www.straptank.com>
16 Badswell Lane
Appleton
Oxford OX13 5JN
UK
Tel: +44 (0)1865 862 382
http://www.straptank.com <http://www.straptank.com/>
Re: change planning
Posted by Noah Slater <ns...@apache.org>.
On Mon, Jan 05, 2009 at 01:26:36PM +1030, Antony Blakey wrote:
> Synchronous discussion such as IRC encourages a single-time-zone of
> interaction.
Maybe in your experience, but not in mine. Me, Jan, Damien, and Chris manage to
coördinate pretty well over disparate time-zones. Sometime's you're discussing
things over breakfast, and sometimes over dinner. Heh.
--
Noah Slater, http://tumbolia.org/nslater
Re: change planning
Posted by Antony Blakey <an...@gmail.com>.
On 05/01/2009, at 1:09 PM, Chris Anderson wrote:
> I think some of this is a symptom of the fact that at least Jan, Noah,
> Damien, and I are on IRC pretty much every day, so it's easy for us to
> think we've got consensus on something without bringing it up on the
> lists. Perhaps the best solution is to push this activity back to the
> lists, with perhaps of list of decisions that have been made (even if
> not implemented) maintained either in the bug tracker or on the wiki,
> with reference to ML threads.
>
> However, all that sounds very formal and slow, and at this point in a
> project with only 5 committers and very simple code base and feature
> set, which is not yet 1.0, I'm inclined to think erring on the side of
> faster/looser may be better.
I think pre-1.0 is the very time that a more transparent operation
should be encouraged, because non-backwards compatible change after
1.0 is considerably more difficult. And given the decree that 0.9 is
the compatibility freeze point, that seems imminent.
Synchronous discussion such as IRC encourages a single-time-zone of
interaction.
Antony Blakey
-------------
CTO, Linkuistics Pty Ltd
Ph: 0438 840 787
One should respect public opinion insofar as is necessary to avoid
starvation and keep out of prison, but anything that goes beyond this
is voluntary submission to an unnecessary tyranny.
-- Bertrand Russell