Reading my last reply to you in the previous whispers thread, I realize I came off as dismissive and vague. That wasn’t my intention, and I apologize.
In addition to the other excellent replies in this thread, I’ll try to shed some light on our product development process. Our general approach to product development is to start off with something simple, get feedback from our users, and use that feedback to rapidly iterate on features to gradually arrive at the best possible user experiences.
In order to learn from our users as much as possible as quickly as possible, this often means making short-term tradeoffs. In the case of whispers, this tradeoff appears as a tension between what’s easiest for third-party developers and what lets us iterate as quickly as possible. We made sure to implement whispers as a feature that third-party developers can still build on top of (and there are already several bots which utilize them!), though the implementation may not be as convenient as possible.
For instance if we were to use NOTICE for whispers, we’d have had to implement a custom tag or something similar to differentiate a whisper from a regular NOTICE (which we use for slowmode notifications and such), which means our clients would need additional, messier logic, which ultimately ends up with us moving slower.
As our products mature, we’re able to increasingly spend time in optimizing our API for third-party developers. This, for instance, is what we’ve finally begun to do with our IRC protocol after two years of focusing on reliability and general chat improvements. We’ll get there with our other features as well, but it takes time and tradeoffs along the way.