Move metadef commands to sub-commands

[Stuart McLaren]

There are a lot of metadef commands shown with the help output:

$ glance |grep md
    md-namespace-create
    md-namespace-delete
    md-namespace-import
    md-namespace-list List metadata definitions namespaces.
    md-namespace-objects-delete
    md-namespace-properties-delete
    md-namespace-resource-type-list
    md-namespace-show Describe a specific metadata definitions namespace.
    md-namespace-tags-delete
    md-namespace-update
    md-object-create Create a new metadata definitions object inside a
    md-object-delete Delete a specific metadata definitions object inside a
    md-object-list List metadata definitions objects inside a specific
    md-object-property-show
    md-object-show Describe a specific metadata definitions object inside
    md-object-update Update metadata definitions object inside a namespace.
    md-property-create Create a new metadata definitions property inside a
    md-property-delete Delete a specific metadata definitions property inside
    md-property-list List metadata definitions properties inside a specific
    md-property-show Describe a specific metadata definitions property
    md-property-update Update metadata definitions property inside a
    md-resource-type-associate
    md-resource-type-deassociate
    md-resource-type-list
    md-tag-create Add a new metadata definitions tag inside a namespace.
    md-tag-create-multiple
    md-tag-delete Delete a specific metadata definitions tag inside a
    md-tag-list List metadata definitions tags inside a specific
    md-tag-show Describe a specific metadata definitions tag inside a
    md-tag-update Rename a metadata definitions tag inside a namespace.

We could reduce the number by creating a 'metadef' command and moving these to sub-commands under that command.

[Travis Tripp (travis-tripp)]

I’d probably remove the md- prefix on them as well.

metadef namespace-create not metadef md-namespace-create.

That seems more friendly to me.

Actually aligns with openstack client a bit better as well. (although it would probably be metadef namespace create / metadef tag create / etc)

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/218864

[OpenStack Infra (hudson-openstack)]

Change abandoned by Niall Bunting (<email address hidden>) on branch: master
Review: https://review.openstack.org/218864
Reason: -2 from core reviewer.

[Flavio Percoco (flaper87)]

Why should we treat metadef different that other commands?

Why having a subcommand for metadef and not for other things (say task)?

I think this change would make glanceclient's CLI inconsistent, therefore I'm inclined to -1 the proposal

(tagged as spec-lite)

[Stuart McLaren (stuart-mclaren)]

> Why should we treat metadef different that other commands?

The thinking here is that the md- commands are really only going to be used by a very small number of users.

The 99% of users who want to do something with images have to trawl through the 25 md-xxx commands to find the (smaller number) of image related commands they are most likely to be looking for. I don't think this can be considered user friendly.

> I think this change would make glanceclient's CLI inconsistent,

This is a fairly common pattern in UX. Eg firefox has the "back" button always visible, but if you want to do some of the less common actions you drill down into a sub menu. Does this make firefox inconsistent? Or more usable?

[Niall Bunting (niall-bunting)]

Glance IRC comments about this spec

http://eavesdrop.openstack.org/meetings/glance/2015/glance.2015-11-26-14.06.log.html

Related Abstracts:
14:44:14 <openstack> Launchpad bug 1489534 in python-glanceclient "Move metadef commands to sub-commands" [Wishlist,In progress] - Assigned to Niall Bunting (niall-bunting)
14:44:37 <bunting> Just bring this spec lite into the light for moving metadefs into a subset of commands
14:45:09 <nikhil_k> flaper87: anyone sourcing the master HEAD for CI/CD will have breaking changes in their repo (if that's something cared for) (imdone)
14:45:20 <Jokke_> question, I think I have raised before, why only metadefs?
14:45:50 <kragniz> Jokke_: because there's a lot of them
14:45:58 <mclaren> Jokke_: because there's so many md commands -- have you looked at the v2 help? They kind of spam everything else
14:46:28 <flaper87> my current feeling about that is that we should either do it for everything or not doing it.
14:46:30 <mclaren> AS Brian said, it makes glance look like a metadata service with some extra image related bits
14:46:43 <flaper87> if you look at `neutron --help` you'll get a huge list of commands
14:46:47 <mclaren> where really most folks think of it the other way around
14:46:57 <mclaren> flaper87: yes, but they're all networking related
14:47:03 <flaper87> mclaren: right
14:47:13 <kragniz> flaper87: I don't think we want to change general usage of the client
14:47:26 <kragniz> otherwise we might as well just tell people to use osc
14:47:28 <flaper87> kragniz: agreed but this change is a breaking change
14:47:37 <flaper87> I mean, we can proxy the old way to the new way
14:47:40 <mclaren> kragniz: it's backwards compatible
14:47:49 <Jokke_> I'm all up for subcommanding stuff, but I do agree that it would make sense to do it once for all then
14:47:51 <flaper87> but it'll require a major release to completely remove the old one
14:48:04 <mclaren> it's fully backwards compatible
14:48:09 <kragniz> flaper87: that's true, but md-* are far less likely to be used by some end user
14:48:19 <Jokke_> flaper87: that's the beauty of it, all the old ones works while they don't show up
14:49:01 <nikhil_k> so are doing some redirect or something?
14:49:14 <nikhil_k> won't the CLI change (backward incompat)?
14:49:17 <flaper87> I'm not opposed to it, I just shout my first thought
14:49:30 <Jokke_> nikhil_k: the old commands are staying there, just suppressed form help
14:49:50 <flaper87> and I guess we need to keep in mind to remove those on the next major
14:49:51 <Jokke_> making it super easy to transform
14:50:01 <Jokke_> flaper87: or not ;)
14:50:13 <flaper87> Jokke_: the old ones
14:50:18 <nikhil_k> ah, nice trick. So suppressed is fine w/o deprecation (clarification needed)?
14:50:21 <Jokke_> the current trend seems to be bloat all so we can just leave them there
14:50:31 <flaper87> mmh, I don't like having 2 ways of doing things
14:50:31 <nikhil_k> flaper87: ok
14:50:46 <nikhil_k> if we can do redirects , that would be ideal
14:50:53 <mclaren> we have two apis, I thought it was a requirement!
14:51:08 <nikhil_k> (or in bash semantics it would be alias)
14:51:33 <Jokke_> remember ... never remove anything
14:5...

[OpenStack Infra (hudson-openstack)]

Change abandoned by Sean McGinnis (<email address hidden>) on branch: master
Review: https://review.opendev.org/218864
Reason: Very old patch at this point. Feel free to restore and update if you wish to continue with it.