[erlang-questions] next(Erlang): "Fix up records so that records are first-class citizens and not just syntax saccharine for tuples?"

[Richard O'Keefe]

On 22 Nov 2008, at 1:26 pm, Bob Ippolito wrote:
>> Not dismissing it at all. Richard's frames proposal uses atoms for
>> "property names". Every atom can be converted to a string, and every
>> string can be converted to an atom. An atom is basically an interned
>> string.
>
> That's not strictly true, atoms are limited to 255 characters I think.

There is an EEP about that.
As it happens, I wrote it (:-).
To put it briefly, I propose having two representations for
atoms, one of which is basically what we have now, and the
other of which does not have any particular limit on size
or numbers and does not use a centralised symbol table.
In real life, I expect that the limited size symbol table is
a more serious threat to JSON labels -> atom mapping; it is
too easy to break.  This applies to practically *any* use of
atoms in data from sources you do not fully control, which is
why regardless of whether frames/structs are ever adopted or
not, we need better atoms.