# Images of sections of a sheaf

I'm currently reading a paper by X. Caicedo containing an introduction to sheaves. On page 8 he claims, that for every sheaf of sets $p:E\to X$ and every section $\sigma:U\to E$ (U being open in X) the image $\sigma(U)$ is open. This statement is proved by picking a point $e\in\sigma(U)$, an open neighborhood S of e, which satisfies

1. $p(S)$ is open in X,
2. $p\restriction S$ is a homeomorphism

and arriving at an open set $\sigma(U)\supseteq S\cap\sigma(U)=p^{-1}(p(S)\cap U)$.

I think the "$\supseteq$" part of this equation does not hold, if for example E is equipped with the discrete topology and the stalk of $p(e)$ has more than one element.

I have tried to show that $(p\restriction S)^{-1}(p(S)\cap U) = p^{-1}(U)\cap S$ is contained in $\sigma(U)$, but all attempts at that felt quite clumsy, leading me to believe I have missed something important about the structure of a sheaf.

-

The inclusion $\sigma(U)\supseteq S\cap\sigma(U)$ is trivially true.
The equality $S\cap\sigma(U)=p^{-1}(p(S)\cap U)$ is false in general.
For a counterexample take any $X$ and $$E=X\times \lbrace 1,2\rbrace \:,p(x,i)=x\:, U=X\:, \sigma (x)=(x,1)\:, S=X\times \lbrace 1\rbrace$$ where $\lbrace 1,2\rbrace$ has the discrete topology.
You then have $S\cap\sigma(U)=S\neq p^{-1}(p(S)\cap U)=p^{-1}(X)=E$