Exact differential equations and holonomic constraints I understand that if a constraint equation given on a differential form is exact, that means it is also holonomic since I can find a solution. But there are other types of differential equations, like separable and linear, such that I can find an equation in the form of a holonomic constraint. Why is it that being exact is a condition, rather than just having a solution of the type $f(x,y) = 0$?
 A: *

*Here I would like to mention the notion of a semi-holonomic constraint 
$$ \sum_{j=1}^n a_j(q,t)~ \mathrm{d}q^j + a_t(q,t)~\mathrm{d}t~=~0, \tag{1'}$$
which puts an
inexact differential equal to zero.

*It is possible to incorporate semi-holonomic constraints into Lagrange equations, cf. e.g. my Phys.SE answer here.

*Note that eq. (1') does not mean that we demand that the $n+1$ co-vector components 
$$ a_t(q,t)~= a_1(q,t)~= \ldots ~=~a_n(q,t)~=~0 \qquad\qquad (\longleftarrow \text{Wrong!} )
\tag{2'}$$
of a co-vector (1') should be zero. Perhaps this potential misunderstanding (2') is secretly the core of OP's question? 

*Rather eq. (1') means that
$$ \sum_{j=1}^n a_j(q,t)~ \dot{q}^j + a_t(q,t)~=~0. \tag{3'}$$

*In particular, a holonomic constraint 
$$f(q,t)~=~0 \tag{0}$$ 
can be put on the above semi-holonomic form 
$$ \mathrm{d}f~=~\sum_{j=1}^n \frac{\partial f}{\partial q^j}~ \mathrm{d}q^j +  \frac{\partial f}{\partial t}~\mathrm{d}t~=~0 .\tag{1}$$
Explicitly, eq. (1) means that the total time derivative
$$ \frac{df}{dt}~\equiv~ \sum_{j=1}^n \frac{\partial f}{\partial q^j}~ \dot{q}^j +  \frac{\partial f}{\partial t}~=~0\tag{3}$$
is zero.
References:


*

*H. Goldstein, Classical Mechanics, Section 2.4.

