Let A,B,C be three subgroups of G.
(i) The subgroup A normalizes the subgroup (A,B).
(ii) If the subgroup (B,C) normalizes A, the subgroup (A,(B,C)) is generated by the elements (a,(b,c)) with a∈A,b∈B and c∈C.
(iii) If A,B and C are normal, then
(A,(B,C))⊂(C,(B,A)).(B,(C,A)).
By (4 bis), for a,a′∈A and b∈B,
(a,b)a′=(aa′,b).(a′,b)−1,
whence (i). Suppose now that (B,C) normalizes A. For a∈A,b∈B,c∈C and x∈G, (4) implies
(a,(b,c).x)=(a,x).(x,((b,c),a))(a,(b,c))
and ((b,c),a)∈A since (B,C) normalizes A, whence by induction on p the fact that (a,∏i=1p(bi,ci)), for bi∈B,ci∈C, belongs to the subgroup generated by the elements of the form (a,(b,c)). If finally A,B and C are normal, so are the subgroups (A,(B,C)), (C,(B,A)) and (B,(C,A)). It therefore suffices by (ii) to show that
(a,(b,c))∈(C,(B,A)).(B,(C,A))
for all a∈A,b∈B and c∈C. Now by (5), writing ab−1=u
(a,(b,c))=((ub),(b,c))=(cu,(u,b))−1.(bc,(c,u))−1,
whence (iii).
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.
Nothing.