1cb77081b0
The doc build fails locally because it treats warnings as errors. (In the gate, treating warnings as errors is waiting for [1].) [1] https://review.openstack.org/229951 Change-Id: I84ab0de7a89dca1d53d0a5157d83aff0a1c31d95
79 lines
3.4 KiB
ReStructuredText
79 lines
3.4 KiB
ReStructuredText
=======================
|
|
Neutron-Lib Conventions
|
|
=======================
|
|
|
|
Summary
|
|
-------
|
|
|
|
* Standard modules - current cycle + 2 deprecation policy
|
|
* Legacy modules - current cycle + 1 deprecation policy
|
|
* Private modules - no deprecation warnings of any kind
|
|
|
|
Interface Contract
|
|
------------------
|
|
|
|
The neutron-lib repo exists to provide code with a long-term stable interface
|
|
for subprojects. If we do need to deprecate something, a debtcollector
|
|
warning will be added, the neutron-lib core team will make every effort to
|
|
update any neutron stadium project for you, and you will get at least the
|
|
current release plus two cycles before it disappears.
|
|
|
|
In keeping with how hard it is to remove things, the change velocity of this
|
|
library will be slower than neutron. There are two notable cases where code
|
|
should go into standard neutron instead of this lib (or your repo):
|
|
|
|
* It is something common, but changes a lot. An example is something like
|
|
the neutron Port object. Everyone uses it, but it changes frequently.
|
|
You don't want to wait for a library release to tweak some neutron feature,
|
|
and we are not going to force releases quickly because you tried to put
|
|
it here. Those items will need to be addressed in some other manner
|
|
(in the case of the Port object, it'll be via an auto-magic container
|
|
object that mimics it.)
|
|
|
|
* It is something common, but you need it now. Put it in the repo that needs
|
|
it, get your stuff working. Then consider making it available in the lib,
|
|
and eventually remove it from your repo when the lib is publishing it.
|
|
An example would be a new neutron constant. The process would be, put it
|
|
in neutron along with your change, submit it to the lib, when that constant
|
|
is eventually available, remove it from neutron and use the lib version.
|
|
|
|
Private Interfaces
|
|
------------------
|
|
|
|
Private interfaces are *PRIVATE*. They will disappear, be renamed, and
|
|
absolutely no regard will be given to anyone that is using them. They are
|
|
for internal library use only.
|
|
|
|
DO NOT USE THEM. THEY WILL CHANGE.
|
|
|
|
Private interfaces in this library will always have a leading underscore,
|
|
on the module or function name.
|
|
|
|
Legacy Modules
|
|
--------------
|
|
|
|
This library has a special namespace called neutron_lib.legacy.
|
|
|
|
Anything in this directory will likely get a new interface in the top-level
|
|
library sometime in the near future, and then a debtcollector deprecation
|
|
notice. Expect to get current cycle plus one release of maintenance at that
|
|
point, and then they will be removed.
|
|
|
|
Why this intermediary step? Because neutron has some serious dependency
|
|
issues with its subprojects that need breaking, we do not want to rush
|
|
some of the refactors to our interfaces that need to happen, we have
|
|
limited resources, but we still need to make addressing those dependency
|
|
issues a high priority.
|
|
|
|
The legacy module is for those existing modules in neutron that are in
|
|
wide use by subprojects, but which are not super interfaces. The legacy
|
|
submodule is for routines that will still be maintained with a long-term
|
|
backwards compatibility interface contract, but which are not considered
|
|
"library worthy" by the neutron core team.
|
|
|
|
This can easily be abused as a kitchen sink to just move stuff and make
|
|
fast progress. Please do not do this, and do not expect this kind of thing
|
|
to be favorably reviewed. Good candidates for this area are things that
|
|
we want to refactor, but are lower priority, AND they have been around for
|
|
a long time with no changes (i.e. an existing history of stability).
|