Re: scratch/command 064f146 1/2: Change command to interactive ... modes

[Óscar Fuentes]

Juri Linkov <juri@linkov.net> writes:

>>>>> As for verbosity, it's not like there weren't any suggestions to
>>>>> reduce the number of annotations either (like tying command
>>>>> definitions to packages, or custom groups, etc).
>>>>
>>>> And I responded to all of those and said why I thought those were bad
>>>> ideas.
>>>
>>> Then why we are removing :group tags from defcustom?  Because they are 
>>> redundant.
>>> Completion annotations are redundant the same way.
>>
>> Not exactly in the same way, because the overwhelmingly most common case
>> with :group tags is that all defcustoms in one file share the same group.
>>
>> With completion annotations, we have seen ratios from 50-75 %.
>
> I don't know where you see such low ratios.  In a typical package like
> gomoku.el there are 16 commands tagged with 'gomoku-mode', and 2 untagged.
> This is 90% vs 10%.
>
> It makes more sense to make tagging opt-out, i.e. to tag the whole package
> like defgroup does.  Then tag only a few commands available globally.
>
> Usually most commands in a package are internal, only a few of commands are
> entry points.  Package entry points are usually tagged with the autoload 
> cookie.
>
> So only 10% of entry points need the `declare' tag, other 90% don't need any 
> tagging.

As mentioned elsewhere, tagging is far from being a simple task. Each
command must be considered individually. Your example about gomoku is
not representative as it is an small, isolated package, i.e. you won't
use gomoku-* elsewhere outside gomoku, except for the entry point(s), of
course.

The statistics Lars posted about Gnus are telling: only about 6 of every
10 commands were tagged. I expect to see similar numbers on other large
packages, such as Org and Calc.