"On the receive side, you can make your receiver as narrow at the transmitters oscillator". Completely wrong, and shows a fundamental misunderstanding.
Morse is a wonderful mode, but suffers from the lack of error correction which had made modes like PSK31 and FT8 so much more effective, especially in bandwidth usage.
Clrly yu hve never usd this on the bands.
Yes, the error correction is between the operator ears. It is not a mode that can easily automated.
CW is wonderful, but is has long since been superseded by more modern digital modes.
You can easily use a transmitter with Gaussian keying to reduce the sidebands to zero.
The sidebands in CW ARE the information.
> When the key is down there is no modulation whatsoever.
No, but you can't keep the key down forever. The information is in the transition from Key-down to Key-up (and Vice/versa). Which is why the modulation envelope has to be tailored to the keying speed, else you get excessive key clicks.
> You can easily use a transmitter with Gaussian keying to reduce the sidebands to zero.
Gausian keying requires a Raised-Cosine envelope. It is this which contains the sidebands and the information, and causes the B/W to be widened.
Morse is the ultimate narrowband mode. The cleaner your oscillator and the narrower your output the better. On the receive side, you can make your receiver as narrow at the transmitters oscillator.
The only comparable digital mode is on-off keying which is used by cheap, low data rate devices. Even 'narrow' digital modes are wideband compared to morse code.