bug#33225: [debbugs.el] Don't send control message immediately

[Michael Albinus]

"Garreau, Alexandre" <galex-713@galex-713.eu> writes:

Hi Alexandre,

>> As mentioned in [1], I have a modified version of
>> debbugs-gnu-send-control-message which just creates the message (or
>> edits the current one) rather than sending it right away.
>
> I believe this is the most straightforward solution to the problem of
> control messages sending too fast for the newcomer user.

As you have seen the other message, I'm also in favor of this solution.

> However, I guess this is a feature more than a bug for most used
> users, and this would have been only useful for the first time to me,
> when I wanted to “test it to see how it looks” thinking of “cancelling
> at the last moment”, and discovered the mail was already sent before I
> could even understand what was happening (and take the time to
> document more myself on what was the control message really meaning).
> Sure if I saw a message-mode buffer before, at this moment, I’d have
> canceled, but that would only have been useful the first time: then it
> would have become tiring, so I would have disabled it to come back the
> simple behavior, and loose the benefit of the additional
> security-check step.

In fact I believe we shall offer both
functions. debbugs-gnu-send-control-message, without any confirmation,
is good for experienced users. When you are working on a bug triage, for
example, you need a fluid workflow, w/o interruption for confirmation
and alike.

> What I suggested later instead was asking for confirmation for certain
> control messages that might uselessly garble and bloat the bug history,
> while letting go more common and useful messages that might otherwise
> anyway be undone without semantic overhead [2], first suggestion was
> just about asking for confirmation about certain messages and not
> others, so I attach the already suggested [3] patch here for better
> referencing of it.  But in last message [2] I argued in favor of better
> discrimination among control messages and potentially clarifying their
> meaning and better linking doc.

I believe it is a matter of personal taste, which control messages need
a confirmation, and which not. And this will change also over the
time. Oce you know what "owner" is intended for, you don't need to
confirm this again and again.

What I could imagine is, that you implement an explanation for the
control messages, based on Noam's debbugs-control-make-message. This
needs an emhancement in the manuals, but I'm working on.

Best regards, Michael.