This is a closing remark in a paper of Elon Lima, Separation Theorem for Smooth Hypersurfaces.
He proves a well known lemma: Suppose $M\subset\mathbb{R}^m$ is a compact, orientable, smooth hypersuface. Fix a smooth field of unit normal vectors $v:M\to\mathbb{R}^m$, and define a smooth map $h:M\times\mathbb{R}\to\mathbb{R}^m$ by $h(x,t)=x+t\cdot v(x)$. Then for some $\epsilon>0$, $h:M\times(-\epsilon,\epsilon)\to\mathbb{R}^m$ is a diffeomorphism onto an open subset of $\mathbb{R}^m$.
He later remarks that if $M$ is only known to be closed, the lemma can be modified to change $\epsilon$ from a constant to a continuous positive function $\epsilon: M\to\mathbb{R}$ with the property if $x|y$ are in $M$ then $x+s\cdot v(x)|y+t\cdot v(y)$ for all $s\in(-\epsilon(x),\epsilon(x))$ and $t\in(-\epsilon(y),\epsilon(y))$.
I'm just a layman, but is there a pedestrian proof on why such a function $\epsilon$ exists? Also, what does the notation $x|y$ mean? Thanks!
NB: I posted this on MSE a couple days ago, but didn't get much of a response.