A is true.
One possible variable-record representation is:
$$\mathrm{F1\ \$\ F2\ \$\ F3\ \$\ F4}$$
but then locating a later field requires scanning through delimiters.
We can keep field offsets in the record header, so delimiters are unnecessary.
B is false.
A record header itself consumes metadata space.
Therefore, it is not guaranteed to use less storage than every possible fixed-length representation.
C is true.
Suppose the header contains offsets:
$\mathrm{offset(F1)}$
$\mathrm{offset(F2)}$
$\mathrm{offset(F3)}$
$\mathrm{offset(F4)}$
The DBMS can obtain the beginning of $\mathrm{F3}$ directly rather than scanning $\mathrm{F1}$ and $\mathrm{F2}$.
D is true.
NULL can be represented through header metadata or a null pointer rather than reserving the full field value space.
