i don't remember the answer to your question, but I suspect it is this:

# 1,2 ;;
- : int * int = (1, 2)

That is, again, syntactic succinctness.  tuples need not always be
parenthesized.  But I'm really just making this up -- I certainly don't
remember well enough to be definitive on this.

It does seem to follow the general pattern, though: with generativity of
record-field-names as another example.

On Wed, Feb 14, 2018 at 3:40 PM, Ian Zimmerman <itz@very.loosely.org> wrote:

> On 2018-02-14 15:02, Chet Murthy wrote:
>
> > I remember back in the day Pierre Weis explaining to me that this
> > syntactic trade-off was made in order to allow that "let" and "match"
> > didn't have ending key-words (e.g. "end").
>
> I would understand and accept that choice.  But the larger question is,
> why was the semicolon overloaded like this?  In SML the semicolon serves
> just one purpose: separating consecutive imperative statements (well it
> can also separate declarations but that is optional).  Why has CAML
> chosen to use the semicolon in list and record patterns and values,
> instead of the comma like SML?
>
> > Unlike in SML/NJ.
>
> SML has multiple implementations (as implied in the S).
>