On 8/21/24 5:50 PM, Jim Fehlig via Devel wrote:
Laine's attempt long ago [1] to deprecate/obsolete the virInterface*
APIs did not receive a standing ovation. However he raised many good
points which are still valid today. If anything, netcf, the libvirt
netcf backend, and the whole interface driver have become more stale.
Personally, I wish we could deprecate and eventually obsolete the APIs
and nuke the entire driver, but understand that's not
desirable/realistic upstream. Given the state of the thing, I'd like to
hear opinions on adding some words to the interface API page and
virtinterfaced man page that describe the state of the APIs/driver and
discourages their use.
The motivation for this request comes from my efforts to take advantage
of libvirt's modularization and not provide virtinterfaced
(libvirt-daemon-driver-interface package) for certain use-cases. Having
upstream documentation warning of the limitations and inactive
development of the APIs/driver would help in convincing unsuspected
users to find a better solution. And I think the use of "unsuspected" is
correct :-). Projects closer to libvirt, like virt-manager, ditched use
of the APIs long ago. But projects with less cross-pollination may not
be aware of the general uselessness.
As an example, we recently discovered cockpit-machines uses
virConnectListAllInterfaces and virInterfaceGetXMLDesc. It's not clear
to me why cockpit gathers this info. AFAICT it's not used for anything.
The code was added long ago [2] with no info on purpose or what it's
supposed to achieve. We've opened a github issue [3] to discuss removing
use of these APIs. Pointing cockpit-machines developers to upstream
libvirt documentation that describes the state of virInterface* and sets
realistic expectation might help in convincing them to follow
virt-managers footsteps and drop use of the APIs.
From the looks of the issue you linked, it seems Martin Pitt has now
removed all that.
I can provide the words of discouragement in the form of a patch if
folks are receptive to the idea :-).
I would obviously not complain, but there may be other opinions :-)
Regards,
Jim
[1]
https://listman.redhat.com/archives/libvir-list/2020-December/212791.html
[2] https://github.com/cockpit-project/cockpit/pull/12465
[3] https://github.com/cockpit-project/cockpit-machines/issues/1777
