[petsc-dev] petsc-dev on bitbucket

[Satish Balay]

On Thu, 9 Feb 2012, Jed Brown wrote:

> On Thu, Feb 9, 2012 at 23:50, Satish Balay <balay at mcs.anl.gov> wrote:
> 
> > On Thu, 9 Feb 2012, Jed Brown wrote:
> >
> > > On Thu, Feb 9, 2012 at 23:36, Satish Balay <balay at mcs.anl.gov> wrote:
> > >
> > > > > > Credential management is simpler because adding/removing keys and
> > > > similar
> > > > > > can be done by the user instead of by emailing Satish.
> > > > >
> > > > > replaced by  'crate a bitbucket account - and email sean your
> > bitbucket
> > > > id'
> > > > >
> > > > > so eqivalent.
> > > >
> > > > ok - I see your point. First step is equivalent. Later change of
> > > > keys/passwd is up to the user and doesn't involve e-mail [ with
> > > > bitbucket.]
> > >
> > > exactly
> >
> > Just want to mention: this model [each user have their own account,
> > and access shared repos] is the historical model with vcs [sharinv via
> > unix group permissions]
> >
> 
> This is actually a quite different model.
> 
> 
> >
> > And we migrated to ssh key management as superior -and abandoned the
> > previous model. Firstly because the repos were at mcs - and accounts
> > were not easy for non mcs folks - and then not dealing with groups was
> > nicer. And when petsc.cs.iit as deployed [eventhough I could easily
> > deal with user accounts and groups]- it was preferable to stick with
> > ssh key model instead of user account model. [with the aversion we
> > gradually developed for maintaing multiple accounts/multiple paswds
> > all over].
> >
> 
> The permissions issues with this are a mess.
> 
> 
> >
> > Now we are back to the original model of user account - and access via
> > this account [yes - bitbucket makes it easier to crate accounts, and
> > hopefully better infrastrucutre than 'group permission' for sharing
> > underneath. [well atleat there is a gui to it].
> >
> 
> With Bitbucket, GitHub, and all the others, the web account is *not* a unix
> account. It's just a way to manage identity/keys. Everyone using ssh has
> hg at bitbucket.org (and git at github.com, etc).

sure the implemtation details are different - but I mean - the top
level models as I described are equivalent - and can be implemented
equivallly.

Yesh all users go though a single unix account [for ssh] - and a
[thats implementation detail - and it shouldnm't matter].


satish