role/internal is require in order to have loggable users.

Vifib uses erp5_credential and have to be configured.

Users coming from web shall be able to register to vifib web site and shall be
able to request functional Software Instance.

remove vifib_credential because this bt5 is not required anymore. all scripts were moved to erp5_credential and refactored to be generic

If buildout binary is not passed the next step is try to import zc.buildout
before falling back to internal bootstrap.

Thank to this it will be possible to use system or buildout provided buildout
with all related dependencies (like slapos.libnetworkcache).

Fetch slapos.libnetworkcache, as will be used sooner or later.

Pin newer buildout.

This reverts commit 4ac582c4.

This reverts commit f2be5210.

The idea *was* to not force providing external buildout. But then slapgrid
would have to get all not known dependencies, so it will be less usable.

Drop this way.

This reverts commit f2be5210.

The idea *was* to not force providing external buildout. But then slapgrid
would have to get all not known dependencies, so it will be less usable.

Drop this way.

Use the best currently known version and select buildout location ouside of
pypi.

In slapos environment more flexibility on bootstrap process is required, and
slapos.buildout allows to select non-pypi location of zc.buildout distribution,
independently from setuptools/distribute location.

Old version of slap library are not using state parameter, so prove that new
version of Slap Tool is capable to speak with them.

Do not try to request start not fully installed computer partition, and wait
until instance setup packing list is stopped, before trying to create hosting
one.

Beside of setting state on Software instance by start/stop requested methods
setup in activities real request. Thanks to this *after* computer partition is
installed (instance setup packing list == stopped) the request becomes
synchronous (...a bit later -- in sense -- just once called).

It shall check for instance hosting, not instance setup.

Setup all required parameters before requesting again.

If internal bootstrap is used there is no need to clutter arguments with not
required command.

Play usual started scenario, and later after it is started, change the decision
and request it stopped.

Also fix docstring.

This was not implemented, and I overlooked it when implementing
-d/--dry-run option for some reason.

It is required in order to pass values like None.

Keep backward compatibility and start automatically just after partition become
available.

Implement start_requested and stop_requested states in order to have stateful
information about software instance requested state.

Asynchronously read software instance state and take proper actions in such
case (only after .available() call for now).

Note: Done in one big commit, as changes are interrelated, and impact system
functionality.