Re: [PATCH 0/2] Replace anti-social QOM type names (again)

[Markus Armbruster]

Thomas Huth <thuth@redhat.com> writes:

> On 14/11/2023 08.41, Markus Armbruster wrote:
>> Cc: the other QOM maintainers
>> Daniel P. Berrangé <berrange@redhat.com> writes:
>> 
>>> On Mon, Nov 13, 2023 at 02:43:42PM +0100, Markus Armbruster wrote:
>>>> We got rid of QOM type names containing ',' in 6.0, but some have
>>>> crept back in.  Replace them just like we did in 6.0.
>>>
>>> It is practical to add
>>>
>>>     assert(strchr(name, ',') == NULL)
>>>
>>> to some place in QOM to stop them coming back yet again ?
>>
>> This adds a naming rule to QOM.  Right now, QOM has none whatsoever,
>> which I've long called out as a mistake.
>>
>> I'm all for correcting that mistake, but I'd go further than just
>> outlawing ','.
>
> What prevents us from fixing this "mistake"?

1. Having to clean up the naming messes we made.  This involves backward
compatibility arguments and work-arounds.

2. Inertia.

> Is there any compelling reason for keeping the current lax naming rules of 
> QOM?

Can't think of any but avoiding 1.

> Would there be migration issues if we'd rename the current offenders? (and 
> even if so, couldn't we simply fix that issue by curating an allowlist of old 
> names?)

I believe migration should not be affected, since migration section
names are entirely separate.  Mind, I'm no migration expert.