ierfc(z) : sign definition

[Gil]

Function ierfc(z)

Wolfram Alpha defines it as
z×erfc(z) - e^(-z²)/sqrt(pi)
and calculates accordingly.

At Wikipedia, French version,
of Function d'erreur, section "Extensions",
we find the reverse definition, ie:
e^(-z²)/sqrt(pi) - z×erfc(z).

Is there an "error" here?

What is the most common convention or the "correct" definition ?

Thanks for your answer.