Alexander's subbase theorem There are two different definitions for a subbase of a topological space that are not equivalent:
Let $(X,\mathcal{T})$ be a topological space.
1) $\gamma \subset \mathcal{T}$  is a subbase of $\mathcal{T}$ if the collection of all finite intersections of elements of $\gamma$ together with the set $X$, forms a basis for $\mathcal{T}$.
2) $\gamma \subset \mathcal{T}$ is a subbase of $\mathcal{T}$ if additionally $\gamma$ covers $X$, i.e. $\bigcup\limits_{U \in \gamma} U=X$.
theorem of alexander: If every open cover by elements from $\gamma$ has a finite subcover, then the space is compact.
question: Does the theorem holds even for the 1) defintion?
considerations: If $\gamma$ doesn't cover $X$, then for each open cover $\mathcal{U}\subset \mathcal{T}$ of $X$ there holds $X \in \mathcal{U}$ (since each open set $U \in \mathcal{U}$ can be expressed as union of finite many intersections of elements of $\gamma$) and hence $X$ is a finite subcover of itself. So $X$ is compact. 
 A: Yes because with the addition of X, $\gamma$ will cover X.
The 2nd definition is much preferred to the unusual, weaker 1st definition. 
A: In my definition a subbase $\mathcal{S}$ for a topology $\mathcal{T}$ is a subfamily of $\mathcal{T}$ such that the smallest topology on $X$ that contains $\mathcal{S}$ equals $\mathcal{T}$. 
So $\emptyset$ is actually a subbase for the indiscrete (trivial) topology on $X$ in that definition. 
So a subbase for $\mathcal{T}$ need not contain a cover of $X$, as Munkres demands in his text (it is the only popular textbook that defines a subbase this way), but what is true is that a base for $\mathcal{T}$ is given by all finite intersections from $\mathcal{S}$ under the proviso that the empty intersection (the intersection of $0$ members of $\mathcal{S}$ equals $X$, the void intersection "convention") and of course the intersecton of a one set family is that one set in the family etc. All these count as finite intersections)
Alexander's subbase lemma holds both for my definition of subbase as well as the Munkres one. You can always just add $X$ for free to any of "my" subbases and get a valid "Munkres" subbase. This adds no new interesting covers to check.
