This is again much a question about terminology, without only history
as a possible authority to determine "what is right".
These two kinds of values are very similar in the sense that they
should be considered as atomic entities, distinguished only by their
representation as a string of character over some alphabet, possibly
with some minor constraints regarding the use of some characters (e.g.,
should start with alphabetic character,). These constraint are really
immaterial, since what matters mainly is whether two symbols are the
same or are different.
A major difference is that the set of symbol is potentially infinite
(like the set of strings meeting the constraints) though actually
limited by the implementation of the language, like anything else.
An enumerated type, is defined by giving explicitly an enumeration of
the values of that type, which also have a representation as a string
of characters. Since the enumeration is explicit, there is only a
finite number of values in an enumeration type. Languages often consider that
these values are ordered in by the order of the definition. Thus these enumerated
value can be compared for equality, of for order.
Symbols can only be compared for equality. Sometimes, an arbitrary (alphabetical ?) ordering may be provided as it may help for some algorithms. But there is no semantics attached to it.
In a language, You usually have only one type for symbols. However, it
is easy to make more when you have data abstraction available. Symbols
are usually implemented like identifier tables in interpreted
languages, and can actually stand for identifiers in meta-circular
interpreters.
Enumerated types are defined by the programmer, and he may define
several such types. One source of problems is that he may (at least in some languages)
use the same name for different enumerated values, not in the same
type of course, but in different enumerated types.
type month = (January, February, March, April, ...)
type girl = (Mary, April, Nancy, Ann, ...)
type city = (London, Paris, Nancy, Roma, ...)
type writer = (Hemingway, London, ...)
Of course, I am abbreviating illegally, since the list must be
completely enumerated.
But if I use the identifier Nancy in the program, I may not know
which is meant, the girl or the city. This can be resolved by various
means (scoping, casting, etc.)
You do not usually get this problem with symbols, as there is only one
type and a specific syntax for its values.
If you can define more by abstraction, then you usually have to use
different syntaxes for the values. But this is getting into other
issues.
For more specific points.
Your quote would be more accurate if it said "Symbol is a primitive
datatype ..." A symbol is a value in that datatype.
It is likely that some people will talk of symbols for values of an
enumerated type. Another word is token. But these two words have so
many uses... Again, I do not wish to enter a terminology dispute. It
is moderately constructive.
