DFT phase polarity

Just out of curiosity: what is the reason that the FFT in Igor has opposite phase polarity compared to most other programs (as stated in the manual)?
While the polarity is 'merely' a convention, this caused me some confusion comparing spectra between some numpy code and Igor, and it seems like most other numerical programs use
fft[n] = sum_k(data[k] * exp(-2pi*i*k*n/N)) as opposed to Igor's
fft[n] = sum_k(data[k] * exp(2pi*i*k*n/N)).

Cheers,
Fabrizio
Many programs are based on FFTW. I am not sure about the polarity or the reasoning behind the code/algorithm in FFTW.

The code we have used in Igor since ~1995 is based on code that I initially used in grad school. The choice of sign was appropriate for my research applications at the time.

A.G.
WaveMetrics, Inc.