Gerd's reply covers most of the important points. I just wanted to add:

On 16 February 2015 at 00:06, Jean Saint-Remy <jeansaintremy@yahoo.com>
wrote:

> I could not declare the fields for example as a polymorphic tuple for
> instance: "type idref = { mutable 'a * 'b -> 'a }" nor "type idref = {
> mutable 'a. 'b -> 'a }". The latter is just nonsense, the 'b is undefined.
> But we could have a two field record by just repeating the type
> definitions. "let idref = { mutable 'a. 'a -> 'a; mutable 'b. 'b -> 'b }".
> The polymorphic type does not have any restriction, it appears we could
> have any fundamental type we wanted. The type 'a variable must be repeated,
> it is not enough to say "type idref = { mutable 'a. -> 'a };;" Is it
> correct then to read the declaration in this way: the type 'a polymorphic
> type as a polymorphic variable of type 'a. I know it sounds redundant, but
> that is how I am reading the syntax.
>

All the pieces of record syntax in this quoted paragraph are missing a
field name. Where you say things like "type idref = { mutable 'a * 'b -> 'a
}", you probably mean something like "type idref = { mutable fst : 'a 'b .
'a * 'b -> 'a }" -- notice the name of the field, followed by the colon,
which always comes before the type, including any 'a . -style variable
bindings it might start with.