Anyone telling you it can be done some other way is blowing smoke.
Or smoking something.
Our halfwave verticals aluminium do not need counterpoise as they have a autotransformer that results in a perfect R 50 and J0 impedance.
An excellent public announcement that you really do not understand how your antennas work.
The purpose of the counterpoise isn't to reduce ground losses (though it may do so - just not
to the same extent as with shorter verticals.) It isn't to improve the SWR (though it does
affect the SWR / load impedance to be matched.) The purpose is to reduce the common mode
currents on the feedline.
Kirchoff's law says the current flowing into a node must equal the current flowing out of it.
This applies to the antenna feedpoint: if you have 140mA flowing into the antenna, there
must be 140mA flowing somewhere else. If you had a perfect choke that would prevent
this from flowing on the feedline or anywhere else, then you wouldn't be able to get any
power into the radiator - it would be like having a broken wire on one side of a power cord.
In most cases the coax braid acts as a counterpoise / radial. Because of the high impedance,
almost any choke balun will NOT be effective to prevent this. That doesn't mean that the
operator will necessarily notice any problems, especially at lower power levels. The lower
the impedance of the counterpoise / radials, the more current flows in them instead of on
the coax shield. How long the radials need to be to reduce the common mode current below
a particular value depends on the power level and the impedance looking back down the
coax shield from the feedpoint.
Remember that, while the current may be low compared to that flowing at the feedpoint
of a quarter wave vertical, it is still capable of radiating the same power as the antenna
itself if it flows into a high impedance resonant circuit - for example, if you just have a
half wave of coax connected to an isolated radio. The ideal case for minimizing common
mode current is to provide a high impedance on the coax shield AND a much lower
impedance radial / counterpoise / ground connection where most of the current will flow.
So what is the ideal
length? Probably a quarter wavelength if the purpose is to minimize
common mode currents.
What is the minimum length you can get by with
? That depends on your power level,
coax length, shack grounding, etc. In some cases the answer is "none" if the common mode
currents don't cause any problems. Otherwise you'll have to experiment and see what it
is in your specific installation.