You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@any23.apache.org by Lewis John Mcgibbney <le...@gmail.com> on 2013/03/19 20:20:29 UTC
[DISCUSS] Community Conformation Regarding Move to Git for Any23
Hi All,
I want to discuss the proposal to shortly hold a VOTE on the migration to
Git for Any23. Please read the following
*VOTE*
As you may have noticed, Peter and myself have been pushing to make the
transition to Git for source code storage and management.
Currently the existing Any23 SVN repos [1] is READ only, meaning that we
cannot make the 0.8.0 release. The scheduling for this was mainly my fault
as I opened the initial INFRA ticket [0] asking for migration to take place.
Anyway, we have been advised to look at the proposed Git repos [2] and make
a VOTE as to whether this reflects
1) An accurate depiction of what we currently have within [1]
2) That it includes all other required directories within the root Any23
SVN repos [3]
The VOTE will be solely on the above criteria and will be open for at least
72 hours. I would like to add that it is pretty important (and bordering on
crucial) that we get reasonable feedback on this, also that as many of us
as possible chime in to review/observe/identify discrepancies at this stage
in the migration to Git for Any23. THIS THREAD IS NOT THE VOTE THREAD.
*IMPORTANT OBSERVATIONS* *FOR CONSIDERATION*
Within the existing Any23 SVN repository we have several directories which
are not currently contained within the proposed Git repos. I now account
for why this is the case.
1. /branches - this is empty. So far we have not branched any Any23
codebase to work on as all development has been in trunk.
2. /committers - most importantly this includes the committers KEYS
file, technically meaning that only Simone and myself can push releases at
this stage unless of course others with to add their KEY to this file. Also
the DOAP file is in here and it is important that we keep this up to date.
3. /repo - this seems to be a legacy Any23 repos which we do not need,
but which is doing no harm being in SVN
4. /repo-ext - this is a repos for external dependencies which HAVE NOT
been released yet. The artifact causing the problem here is commons-csv. I
spoke to commons community and they have confirmed that they are nearly
ready to do a release on this library so I am going to head over there and
help push this on so we do not need to hang on to this repos within Any23.
5. /sandbox - this looks like legacy and we DO NOT need it
6. /site - our beloved Any23 site. It is essential that we keep some
aspect of this writable, however I would like to suggest that we keep this
in SVN like trafficserver does [4]
7. /tags - I think we should push the tags to the Git repos
8. This is from the old google code site. We DO NOT need this in Git.
These are the observations I have made, I am very keen to hear your
thoughts and I apologize for the slight cock up in getting things migrated
in a timely, smooth manner. I did not envisage the above occurring.
Thanks
Lewis
[0] https://issues.apache.org/jira/browse/INFRA-5866
[1] http://svn.apache.org/repos/asf/any23/trunk/
[2] https://git-wip-us.apache.org/repos/asf/any23.git
[3] http://svn.apache.org/repos/asf/any23/
[4] http://svn.apache.org/repos/asf/trafficserver/
--
*Lewis*
Re: [DISCUSS] Community Conformation Regarding Move to Git for Any23
Posted by Peter Ansell <an...@gmail.com>.
On 20 March 2013 19:30, Andy Seaborne <an...@apache.org> wrote:
> On 19/03/13 20:29, Peter Ansell wrote:
>>>
>>> 6. /site - our beloved Any23 site. It is essential that we keep some
>>> > aspect of this writable, however I would like to suggest that we
>>> > keep this
>>> > in SVN like trafficserver does [4]
>>
>> Git negative
>> Svn positive : Easiest way to publish website from what I can tell.
>>
>
> I think currently svnpubsub is exactly that - svn.
>
> For Marmotta, we have to have svn for the website - everything else is and
> was from day 1, git.
For me Any23 has been Git from day one (except for those pesky
patches) and it has worked for me so far!
Cheers,
Peter
Re: [DISCUSS] Community Conformation Regarding Move to Git for Any23
Posted by Andy Seaborne <an...@apache.org>.
On 19/03/13 20:29, Peter Ansell wrote:
>> 6. /site - our beloved Any23 site. It is essential that we keep some
>> > aspect of this writable, however I would like to suggest that we keep this
>> > in SVN like trafficserver does [4]
> Git negative
> Svn positive : Easiest way to publish website from what I can tell.
>
I think currently svnpubsub is exactly that - svn.
For Marmotta, we have to have svn for the website - everything else is
and was from day 1, git.
Andy
PS Watching to see how smoothly this transfer goes ...
Re: [DISCUSS] Community Conformation Regarding Move to Git for Any23
Posted by Peter Ansell <an...@gmail.com>.
On 20 March 2013 05:20, Lewis John Mcgibbney <le...@gmail.com> wrote:
> Hi All,
>
> I want to discuss the proposal to shortly hold a VOTE on the migration to
> Git for Any23. Please read the following
>
> *VOTE*
> As you may have noticed, Peter and myself have been pushing to make the
> transition to Git for source code storage and management.
> Currently the existing Any23 SVN repos [1] is READ only, meaning that we
> cannot make the 0.8.0 release.
If you want to dry-run the release using a Git repository you should
be okay to do the experiment now without waiting for [2] to become
writable. The easiest way to give others access to the experimental
release would be to fork the any23 repository on GitHub and add a git
remote to [2] to pick out the master branch to base the release branch
on. In my repository the master branch from [2] fits exactly over the
trunk from the Git repository that the GitHub any23 repository is
based on so you won't need to rebase if you want to experiment with a
release unless something else changes.
> The scheduling for this was mainly my fault
> as I opened the initial INFRA ticket [0] asking for migration to take place.
> Anyway, we have been advised to look at the proposed Git repos [2] and make
> a VOTE as to whether this reflects
> 1) An accurate depiction of what we currently have within [1]
The master branch in [2] contains all of the files in [1] from my checks.
> 2) That it includes all other required directories within the root Any23
> SVN repos [3]
See below. I only see the tags as needing to be present in Git, and
they are there as native Git tags.
> The VOTE will be solely on the above criteria and will be open for at least
> 72 hours. I would like to add that it is pretty important (and bordering on
> crucial) that we get reasonable feedback on this, also that as many of us
> as possible chime in to review/observe/identify discrepancies at this stage
> in the migration to Git for Any23. THIS THREAD IS NOT THE VOTE THREAD.
>
> *IMPORTANT OBSERVATIONS* *FOR CONSIDERATION*
> Within the existing Any23 SVN repository we have several directories which
> are not currently contained within the proposed Git repos. I now account
> for why this is the case.
Using notation below to clarify whether a directory will be in Git or
whether it needs to stay in SVN even after the transfer:
> 1. /branches - this is empty. So far we have not branched any Any23
> codebase to work on as all development has been in trunk.
Git negative : As it is empty so it doesn't matter. Branches are going
to be used extensively in Git for peer review before merging changes
back to the master branch.
SVN negative : No reason to keep directory in SVN.
> 2. /committers - most importantly this includes the committers KEYS
> file, technically meaning that only Simone and myself can push releases at
> this stage unless of course others with to add their KEY to this file. Also
> the DOAP file is in here and it is important that we keep this up to date.
Git negative : Is this an SVN specific mechanism for authorising
people to do releases? Can the DOAP file be created in the Git
repository and be useful to the Apache DOAP process?
SVN positive : SVN specific mechanism for releases?
> 3. /repo - this seems to be a legacy Any23 repos which we do not need,
> but which is doing no harm being in SVN
Git negative : No need to have repository in Git if it is in
Svn negative : Repo should be replicated in the main Apache maven
repository/Maven central
> 4. /repo-ext - this is a repos for external dependencies which HAVE NOT
> been released yet. The artifact causing the problem here is commons-csv. I
> spoke to commons community and they have confirmed that they are nearly
> ready to do a release on this library so I am going to head over there and
> help push this on so we do not need to hang on to this repos within Any23.
Git negative : Hopefully we won't need to use this mechanism in the future.
Svn negative : Once commons-csv releases we will not be relying on
this directory
> 5. /sandbox - this looks like legacy and we DO NOT need it
Git negative
Svn negative
> 6. /site - our beloved Any23 site. It is essential that we keep some
> aspect of this writable, however I would like to suggest that we keep this
> in SVN like trafficserver does [4]
Git negative
Svn positive : Easiest way to publish website from what I can tell.
> 7. /tags - I think we should push the tags to the Git repos
Git positive : I checked the git-wip repository out separately to
confirm they are present :
any23-wip$ git tag -l
any23-0.1-snapshot
any23-0.2
any23-0.2.1
any23-0.2.2
any23-0.3.0
any23-0.4.0
any23-0.4.1
any23-0.5.0
any23-0.6.0
any23-0.6.0-M1
any23-0.6.0-M2
any23-0.6.1
any23-0.7.0-incubating
Svn positive : There are direct links to these tags from Maven POM
files, so if we can keep this directory intact it would be useful.
Future releases will reference tags in the Git repository.
> 8. This is from the old google code site. We DO NOT need this in Git.
Git negative
Svn negative
> These are the observations I have made, I am very keen to hear your
> thoughts and I apologize for the slight cock up in getting things migrated
> in a timely, smooth manner. I did not envisage the above occurring.
I added the git-wip repository as a remote to my personal git clone
and its "master" branch contains exactly the same Git commit history
as the "trunk" branch from the previous git repository, and it
contains the latest commits from subversion, so the migration has
occurred as expected. I also checked out the git-wip repository
separately to check that the tags were still in it, and it works as
expected.
> Thanks
>
> Lewis
>
> [0] https://issues.apache.org/jira/browse/INFRA-5866
> [1] http://svn.apache.org/repos/asf/any23/trunk/
> [2] https://git-wip-us.apache.org/repos/asf/any23.git
> [3] http://svn.apache.org/repos/asf/any23/
> [4] http://svn.apache.org/repos/asf/trafficserver/
> --
> *Lewis*