Libguestfs - Oct 2021 - [PATCH 1/9] qapi: New special feature flag "unstable"

Kevin Wolf <kwolf at redhat.com> writes:
> Am 25.10.2021 um 07:25 hat Markus Armbruster geschrieben:
>> By convention, names starting with "x-" are experimental. 
The parts
>> of external interfaces so named may be withdrawn or changed
>> incompatibly in future releases.
>> 
>> Drawback: promoting something from experimental to stable involves a
>> name change.  Client code needs to be updated.
>> 
>> Moreover, the convention is not universally observed:
>> 
>> * QOM type "input-barrier" has properties
"x-origin", "y-origin".
>>   Looks accidental, but it's ABI since 4.2.
>> 
>> * QOM types "memory-backend-file",
"memory-backend-memfd",
>>   "memory-backend-ram", and "memory-backend-epc"
have a property
>>   "x-use-canonical-path-for-ramblock-id" that is documented
to be
>>   stable despite its name.
>> 
>> We could document these exceptions, but documentation helps only
>> humans.  We want to recognize "unstable" in code, like
"deprecated".
>> 
>> Replace the convention by a new special feature flag
"unstable".  It
>> will be recognized by the QAPI generator, like the existing feature
>> flag "deprecated", and unlike regular feature flags.
>> 
>> This commit updates documentation and prepares tests.  The next commit
>> updates the QAPI schema.  The remaining patches update the QAPI
>> generator and wire up -compat policy checking.
>> 
>> Signed-off-by: Markus Armbruster <armbru at redhat.com>
>
> Obviously, replacing the old convention gets rid of the old drawbacks,
> but adds a new one: While using x- makes it very obvious for a human
> user that this is an unstable feature, a feature flag in the schema will
> almost certainly go unnoticed in manual use.
I thought about this, but neglected to put it in writing.  My bad.

Manual use of unstable interfaces is mostly fine.  Human users can adapt
to changing interfaces.  HMP works that way.

Management applications are better off with a feature flag than with a
naming convention we sometimes ignore.

The most potential for trouble is in between: programs that aren't
full-fledged management applications.

If we want to keep "unstable" obvious to the humans who write such
programs, we can continue to require "x-", in addition to the feature
flag.  We pay for it with renames, and the risk of forgetting to rename
in time (which is what got us the awkward stable
"x-use-canonical-path-for-ramblock-id").  Tradeoff.  I chose not to,
but
if y'all think we should...

What we can't do, at least not easily, is to use *only* the "x-"
convention: it is not reliable.  We'd have to add a way to say 'this is
stable even though the name starts with "x-"'.

Am 26.10.2021 um 11:37 hat Markus Armbruster
geschrieben:
> Kevin Wolf <kwolf at redhat.com> writes:
> 
> > Am 25.10.2021 um 07:25 hat Markus Armbruster geschrieben:
> >> By convention, names starting with "x-" are
experimental.  The parts
> >> of external interfaces so named may be withdrawn or changed
> >> incompatibly in future releases.
> >> 
> >> Drawback: promoting something from experimental to stable involves
a
> >> name change.  Client code needs to be updated.
> >> 
> >> Moreover, the convention is not universally observed:
> >> 
> >> * QOM type "input-barrier" has properties
"x-origin", "y-origin".
> >>   Looks accidental, but it's ABI since 4.2.
> >> 
> >> * QOM types "memory-backend-file",
"memory-backend-memfd",
> >>   "memory-backend-ram", and
"memory-backend-epc" have a property
> >>   "x-use-canonical-path-for-ramblock-id" that is
documented to be
> >>   stable despite its name.
> >> 
> >> We could document these exceptions, but documentation helps only
> >> humans.  We want to recognize "unstable" in code, like
"deprecated".
> >> 
> >> Replace the convention by a new special feature flag
"unstable".  It
> >> will be recognized by the QAPI generator, like the existing
feature
> >> flag "deprecated", and unlike regular feature flags.
> >> 
> >> This commit updates documentation and prepares tests.  The next
commit
> >> updates the QAPI schema.  The remaining patches update the QAPI
> >> generator and wire up -compat policy checking.
> >> 
> >> Signed-off-by: Markus Armbruster <armbru at redhat.com>
> >
> > Obviously, replacing the old convention gets rid of the old drawbacks,
> > but adds a new one: While using x- makes it very obvious for a human
> > user that this is an unstable feature, a feature flag in the schema
will
> > almost certainly go unnoticed in manual use.
> 
> I thought about this, but neglected to put it in writing.  My bad.
> 
> Manual use of unstable interfaces is mostly fine.  Human users can adapt
> to changing interfaces.  HMP works that way.
> 
> Management applications are better off with a feature flag than with a
> naming convention we sometimes ignore.
> 
> The most potential for trouble is in between: programs that aren't
> full-fledged management applications.
> 
> If we want to keep "unstable" obvious to the humans who write
such
> programs, we can continue to require "x-", in addition to the
feature
> flag.  We pay for it with renames, and the risk of forgetting to rename
> in time (which is what got us the awkward stable
> "x-use-canonical-path-for-ramblock-id").  Tradeoff.  I chose not
to, but
> if y'all think we should...
Just to clarify, I'm not implying that we should keep it. I'm merely
pointing out that there is a tradeoff that requires us to make a choice.
The decision for one of the options should be explicit rather than just
happening as a side effect. Documenting that it was a conscious decision
is probably best done by adding the reasoning for it to the commit
message.
> What we can't do, at least not easily, is to use *only* the
"x-"
> convention: it is not reliable.  We'd have to add a way to say
'this is
> stable even though the name starts with "x-"'.
No question.

Kevin

On Tue, Oct 26, 2021 at 11:37:19AM +0200, Markus Armbruster
wrote:
> Kevin Wolf <kwolf at redhat.com> writes:
> 
> > Am 25.10.2021 um 07:25 hat Markus Armbruster geschrieben:
> >> By convention, names starting with "x-" are
experimental.  The parts
> >> of external interfaces so named may be withdrawn or changed
> >> incompatibly in future releases.
> >> 
> >> Drawback: promoting something from experimental to stable involves
a
> >> name change.  Client code needs to be updated.
> >> 
> >> Moreover, the convention is not universally observed:
> >> 
> >> * QOM type "input-barrier" has properties
"x-origin", "y-origin".
> >>   Looks accidental, but it's ABI since 4.2.
> >> 
> >> * QOM types "memory-backend-file",
"memory-backend-memfd",
> >>   "memory-backend-ram", and
"memory-backend-epc" have a property
> >>   "x-use-canonical-path-for-ramblock-id" that is
documented to be
> >>   stable despite its name.
> >> 
> >> We could document these exceptions, but documentation helps only
> >> humans.  We want to recognize "unstable" in code, like
"deprecated".
> >> 
> >> Replace the convention by a new special feature flag
"unstable".  It
> >> will be recognized by the QAPI generator, like the existing
feature
> >> flag "deprecated", and unlike regular feature flags.
> >> 
> >> This commit updates documentation and prepares tests.  The next
commit
> >> updates the QAPI schema.  The remaining patches update the QAPI
> >> generator and wire up -compat policy checking.
> >> 
> >> Signed-off-by: Markus Armbruster <armbru at redhat.com>
> >
> > Obviously, replacing the old convention gets rid of the old drawbacks,
> > but adds a new one: While using x- makes it very obvious for a human
> > user that this is an unstable feature, a feature flag in the schema
will
> > almost certainly go unnoticed in manual use.
> 
> I thought about this, but neglected to put it in writing.  My bad.
> 
> Manual use of unstable interfaces is mostly fine.  Human users can adapt
> to changing interfaces.  HMP works that way.
> 
> Management applications are better off with a feature flag than with a
> naming convention we sometimes ignore.
We will sometimes ignore/forget the feature flag too though, so I'm
not convinced there's much difference there.
> If we want to keep "unstable" obvious to the humans who write
such
> programs, we can continue to require "x-", in addition to the
feature
> flag.  We pay for it with renames, and the risk of forgetting to rename
> in time (which is what got us the awkward stable
> "x-use-canonical-path-for-ramblock-id").  Tradeoff.  I chose not
to, but
> if y'all think we should...
IMHO the renames arguably make life easier for mgmt apps, as they
only need to check for the name without the x- prefix, and they
know they won't be accidentally using the fature from an older
QEMU where it was unstable because the older QEMU had a different
name they won't be checking for.

We can just as easily forget to remove the "unstable" feature
flag, as forget to rename.

The plus point about the feature flag is that it is aligned with
the "deprecated" feature flag, and potentially aligned with a
future "insecure" feature flag.

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|