Hi guys,
I quickly read the discussion, and at this point, I don't see an urgency
to change things.
I mean, I think we should have DIR-SHARED *or* DIR-API, but not both,
except that right now, we haven't released the 1.0.0-RC1, the API doco
is frankly in its infancy, so let's not push to hard on something which
is relatively minor.
When I see questions like 'is the API stable enough', I'm just telling
myself 'guy, we'd better move to a RC1 fast before playing around with
names...'
Let's abot this thread right now, and move on. We will be back with this
question later.
On 8/5/11 3:55 PM, Pierre-Arnaud Marcelot wrote:
On 5 août 2011, at 15:17, Alex Karasulu wrote:
On Fri, Aug 5, 2011 at 3:24 PM, Pierre-Arnaud Marcelot<[email protected]>
wrote:
On 5 août 2011, at 13:24, Alex Karasulu wrote:
On Wed, Jul 6, 2011 at 10:19 AM, Pierre-Arnaud Marcelot<[email protected]>
wrote:
On 5 juil. 2011, at 22:53, Alex Karasulu wrote:
On Tue, Jul 5, 2011 at 1:01 PM, Emmanuel Lecharny<[email protected]> wrote:
On 7/5/11 11:47 AM, Pierre-Arnaud Marcelot wrote:
Just to be clear before I make the changes.
Should I merge DIRSHARED into DIRAPI ? or DIRAPI into DIRSHARED?
My choice would be the first option, as Shared is becoming the API.
What's yours?
DIRAPI is way better, IMHO.
I suggest the opposite because of the investment that has gone into
references for DIRSHARED. And at the end of the day it's still shared
stuff across both main projects, studio and the server. There are more
things that will go into this down the line besides LDAP. If we
release later we can release just parts of it instead of the whole
thing: meaning just the ldap api.
We can still restructure but we're going to unsettle some references
we've put even into the code around these issues from the past. If
you're find with doing away with it then I can live with it but we
will lose more.
Hi Alex,
I understand your point but hopefully JIRA is pretty well built and manages to
keep references perfectly.
No arguments there. Jira is just great.
Have a look a recent issue a user created in a wrong JIRA project,
DIRSERVER-1630.
It has been created in the DIRSERVER project but I moved it later to DIRAPI,
since the issue was related to the LDAP API instead.
During the move of the issue JIRA gave a new ID to the issue in the DIRAPI
project, DIRAPI-47.
The old ID is still valid and the JIRA link
https://issues.apache.org/jira/browse/DIRSERVER-1630 now redirects to the new
issue:
https://issues.apache.org/jira/browse/DIRAPI-47
Lastly, in the Activity section of the issue ('All' sub-section selected), the
move has been registered with both origin and destination values (project,
version).
So, as you see, I'm not really sure we're going to loose anything in the
migration...
Yeah I see this np. However my worry is in labeling this Directory TOP level
project as specific to the LDAP API since we're most likely going to have more
protocol API's added in the not so distant future. However if you guys are
thinking of creating yet another project that is a peer of DIRAPI say
DIRKRBAPI, then this might be possible but then we need to be a little more
explicit. Perhaps then DIRLDAPAPI so we can have DIRKRBAPI etc. See where I am
going?
Yeah, but I can't foresee the future.
Nor can I but when in doubt, without an immediate need or urgency it's best not
to act. I'm really not seeing this as anything more than just renaming
something to rename it. I think as a group we need to avoid such things as we
stabilize both in terms of our products, their documentation and our habits. I
advocated more gitter in the past to help us find the best resting points early
when we were forming but now we need to be a bit more conservative with some
gitter in the right place. This move just does not have value, and it churns
things needlessly. Why do it?
As I already explained, the current situation isn't sane since we have two different Jira
projects associated to a single "code project".
From a developer POV it forces us to maintain versions in both projects and
makes it difficult to maintain a clear roadmap and/or release notes for each
version.
From a users POV, it is confusing and users report issues in both projects.
It depends on how we organize the project (in terms of version control, build
and release).
We could have a single project (the current DIRAPI project for example) and
multiple components in it: ldap, kerberos.
Or we could have two different projects instead...
Yep agreed.
The whole change here is between API and SHARED. Not everything is API related
either. It could just have been called COMMON.
I thought we had all agreed that the current 'shared' sub-project is actually
what we call the API.
If there are things in it that are not API related, then we need to create a
separate 'api' trunk and move all API-related code to this new trunk, while
leaving the common (shared) code in the 'shared' trunk.
In that case, maybe the merge of JIRA projects should not happen.
If we have and/or if we are going to have in a very near future (one year or
less) some code (probably common to a number of projects) which is not released
as part of the API, then we should probably keep DIRSHARED as is and dedicate
DIRAPI to the specific API releases.
The only thing that bugs me currently is that we have two Jira projects for a single
"code project" and we should remedy it.
However I may be over doing it with the categorization. I am just stating that
we should just do this once and not have to deal with such a shift again in the
next year when this new API emerges naturally out of our progress. Just trying
to hint at some way to save us all some more management overhead otherwise it's
a no brainer.
The only thing we'd probably want to do is to create new versions in the DIRAPI
project matching all versions of the DIRSHARED project.
Maybe with a prefix, to avoid any misunderstanding.
0.9.19 in the DIRSHARED project would then become shared-0.9.19 in the DIRAPI
project.
This seems to be getting more involved in terms of managing things. Renaming
things is not really going to add all that much value, but it will mix up or
organization changing links that are embedded in certain places referencing
issues.
It is not mandatory, just a thought I had to maintain some kind of basic
history (for the moved items).
Links to the moved issues will continue to work with both forms (old and new
project ID), as I explained above.
There's probably a better way I just want a bit more thought on it because
really there's no serious urgency or am I missing something here?
There's no "real" urgency but it has to be done because it has already been
voted
We started a vote but we're still discussing this. I just want to get on base
community wise on this and make all the necessary points. Really it's not going
to kill us making this move but it's that we are continuously making these
kinds of changes. Our users are going to get confused additively over time.
That's exactly why I launched the discussion and I'm waiting that we all agree
on the correct change before doing anything.
Consensus and community is the key...
Regards,
Pierre-Arnaud
and more importantly because it is really confusing to have two JIRA projects
associated with a single project.
There's more issues inside the DIRAPI than in DIRSHARED plus DIRAPI is less
commonly used since DIRSHARED has been around so much more. This is my whole
point in that if we need to push a merge let's do it by merging DIRAPI since
there's less investment in it. Plus the API concept holds less than the SHARED
and COMMON concepts do. All the code in this area is not just for exposing an
API. It's for code that is common across our major products: Studio and
ApacheDS.
People are contributing bug and improvement reports in both projects and it's
getting complicated to merge things when doing a release because you have to
look at two spaces.
It also involves managing versions for the two projects, etc. etc...
Understood. So then a merge is valid and I'm not disputing this at all. These
are very good reasons to merge these however let's give up on the Jira project
we have the least time and effort investment in both as committers and as users.
That was the whole point of the initial vote...
I totally agree with you, so the question now is what stays and what is merged
away. My arithmetic on this matter deals with investment time and what
identifier best fits our way of managing and releasing projects.
So I'm a +1 on the merger. A +1 on merging DIRAPI into DIRSHARED where DIRAPI
disappears and a -1 on the reverse due to the reasons I stated above. Please
excuse me for beating a dead horse to death on this matter but I sincerely feel
that we have significant investment in DIRSHARED over DIRAPI over time, energy
and community familiarity.
Best Regards,
-- Alex
--
Regards,
Cordialement,
Emmanuel Lécharny
www.iktek.com
