www-infrastructure-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alan Cabrera <l...@toolazydogs.com>
Subject Re: Brooklyn not in http://people.apache.org/committers-by-project.html
Date Thu, 22 Jan 2015 15:04:21 GMT

> On Jan 22, 2015, at 6:57 AM, jan i <jani@apache.org> wrote:
> 
>> On Thursday, January 22, 2015, Alan D. Cabrera <adc@toolazydogs.com> wrote:
>> 
>> 
>>> On Jan 22, 2015, at 4:45 AM, sebb <sebbaz@gmail.com <javascript:;>>
>> wrote:
>>> 
>>> On 22 January 2015 at 12:20, jan i <jani@apache.org <javascript:;>
>> <mailto:jani@apache.org <javascript:;>>> wrote:
>>>> On Thursday, January 22, 2015, sebb <sebbaz@gmail.com <javascript:;>>
>> wrote:
>>>> 
>>>>> On 21 January 2015 at 17:42, Alan D. Cabrera <list@toolazydogs.com
>> <javascript:;>
>>>>> <javascript:;>> wrote:
>>>>>> 
>>>>>>> On Jan 21, 2015, at 1:54 AM, Daniel Gruno <humbedooh@apache.org
>> <javascript:;>
>>>>> <javascript:;>> wrote:
>>>>>>> 
>>>>>>> 
>>>>>>>> On 2015-01-21 09:47, jan i wrote:
>>>>>>>> 
>>>>>>>>> Can we all agree that we need to put this into LDAP?
>>>>>>>> 
>>>>>>>> I agree with you  that LDAP is the right place for such information,
>>>>> and I
>>>>>>>> do not see so many podlings fail that it should be a major
concern.
>>>>> However
>>>>>>>> I think a discussion on general@i.a.o is needed.
>>>>>>>> 
>>>>>>>> If PPMC/committers maintenance of a podling is moved to LDAP,
the
>>>>> mentors
>>>>>>>> and/or PPMC will no longer be able to maintain it them self.
The
>>>>> script to
>>>>>>>> update LDAP is only available to officers (chairs) and infra,
>> meaning
>>>>> we
>>>>>>>> move a task and the incubator chair needs to agree to that.
>>>>>>>> 
>>>>>>>> If incubator agrees on this approach then I assume David
(v.p.
>> infra)
>>>>> will
>>>>>>>> be relatively easy to convince.
>>>>>>>> 
>>>>>>>> rgds
>>>>>>>> jan i
>>>>>>> Podlings are experiments, not actual projects, in as much as
a
>> podling
>>>>> is no more than a sub project much like mod_ftp or the Nth commons
>>>>> sub-project. Are you suggesting we add LDAP groups for all sub
>> projects? We
>>>>> have LDAP (UNIX) groups for committers for the sake of maintaining a
>>>>> working authorization/authenication scheme, which is a moot point for
>>>>> incubator where we exercise universal commit bit.
>>>>>>> 
>>>>>>> As I understand it, the sentiment seems to be "let's put it in
LDAP
>>>>> instead of tracking it in a file".
>>>>>>> 
>>>>>>> Let's list some pros and cons for this LDAP idea as opposed to
using
>>>>> the auth file for tracking:
>>>>>>> 
>>>>>>> Pros:
>>>>>>> - it's in LDAP instead of a text file.
>>>>>>> - it may (or may not) be easier to get a list of project members
>>>>>>> 
>>>>>>> Cons:
>>>>>>> - Our current setup would be invalidated
>>>>>>> - We'd have to change our account and podling request processes
>>>>>>> - We'd have to change the graduation process significantly
>>>>>>> - The Incubator chair would have to manage all this or we would
have
>> to
>>>>> rework how LDAP works
>>>>>>> - We'd have to create a new OU for this, which would mean yet
more
>> work
>>>>> on all auth schemes
>>>>>>> - There would/could be disputes over what is canonical at graduation
>> if
>>>>> a resolution conflicts with LDAP
>>>>>>> - We would have to import all the previously established podlings
>> into
>>>>> LDAP (this would be no small task)
>>>>>>> - We would likely create precedence for all sub-projects to have
>> their
>>>>> own LDAP group (yet another OU?)
>>>>>>> - Unless someone from the Infra PMC steps up to do this voluntarily,
>> it
>>>>> would have an added cost of $N to do.
>>>>>>> - The auth file would have to have all its current podling auth
>> entries
>>>>> changed to LDAP.
>>>>>>> 
>>>>>>> If the only reason for moving to LDAP is "I won't have to change
a
>> text
>>>>> file", then I really fail to see the reason to do this.
>>>>>>> 
>>>>>>> What is the actual gain here? It certainly won't make anything
easier
>>>>> for infra.
>>>>>>> How does it in any way compare to the cost of doing such a move?
>>>>>> 
>>>>>> I understand why you’re hesitant to take this on.  You manage
>> volunteer
>>>>> staff and have a limited budget for those who are paid.  From your
>> point of
>>>>> view what we have “works” and there are higher priority items that
you
>> are
>>>>> responsible for getting done.
>>>>>> 
>>>>>> With that said, I would point out that your long list of “cons”
is
>>>>> symptomatic of the problem caused by not putting podling committer and
>> PPMC
>>>>> information in LDAP.  The current mechanism of capturing podling group
>>>>> information is disparate, brittle, and inconsistent.  I don’t think
>> that
>>>>> you’re claiming that the current setup is ideal; you just have to deal
>> with
>>>>> limited resources to get things done.
>>>>>> 
>>>>>> What I am proposing is that I construct a plan to get us to a better
>>>>> place.  I will create the conversion scripts, LDIF files, etc. and
>>>>> coordinate the effort with the podlings.  All I need is the assistance
>> of
>>>>> someone w/ the privileges and the experience to review my proposed
>>>>> changes.  I will do the overwhelming bulk of the work.
>>>>>> 
>>>>>> I don’t think there’s anything to lose by this proposal.  If
I don’t
>>>>> follow up w/ what I propose then we are simply left with our current
>> state
>>>>> of affairs.  If I deliver, we’ll be in better shape overall and things
>> will
>>>>> be much more simple and consistent.
>>>>>> 
>>>>>> wdyt?
>>>>> 
>>>>> I think there may be a much simpler solution to this.
>>>>> 
>>>>> The podling committers are already (or can be) added to the svn-auth
>> file.
>>>>> AIUI if  the podling group is not actually used (e.g. for SVN auth)
>>>>> then it is just ignored.
>>>>> 
>>>>> So it seems to me it would be simple to add podling-ppmc groups to the
>>>>> file.
>>>>> This would probably mean a minor tweak to the scripts that generate
>>>>> the people.a.o website, and possibly also to whimsy, but the changes
>>>>> would be relatively minor.
>>>>> 
>>>>> Assuming there is no objection from infra to adding these entries,
>>>>> then this would have the advantage of simplicity (and speed of
>>>>> implementation).
>>>>> 
>>>>> The podling could then easily manage its committer and PPMC lists as
>>>>> they would be in the same place.
>>>>> Much easier to see whether the lists are correct.
>>>> 
>>>> +1 would be nice if the podling status pages could also use this
>>>> information, so PPMC do not need to maintain multiple places.
>>> 
>>> Note that the existing podling status pages don't seem to have a
>>> standard format for the PPMC members.
>>> So updating them from other sources is likely to be tricky. Or indeed
>>> extracting the information.
>>> 
>>> Maybe if that issue were fixed there would be no need to involve
>>> changes to Infra processes or data.
>> 
>> Status files have a use, being able to scrape information form those HTML
>> files are not one of them.  That is a very, very, poor practice.
>> 
>> No, ultimately those files should be automagically be generated from
>> canonical sources of data, removing yet another source of data that needs
>> to be maintained.
> 
> 
> Yes but please do not forget the biggest problem with LDAP remain
> undiscussed. Do we really want to make a system where mentors/PPMC cannot
> make updates, but need to go through IPMC. chair or infra (this is the case
> with ldap and not likely to change).
> 

You bring up a good point. But this is a tooling problem and one that can easily be solved
by wrapping up the management of all these groups in a rest API. This API can be accessed
by a variety of tools including whimsy.

My general efforts here are not simple busywork. Disparate sources of data, often wildly inconsistent,
need to be consolidated. The consolidated, canonical, data sources need to be put behind rest
APIs where they can be accessed by a variety of tooling, including whimsy. Once we get to
this state all manner of automation is possible. 


Regards,
Alan



Mime
View raw message