>> `condition-ref` is just like `foreign-error-ref` and the field system
>> works just like the 198 property system. However, each SRFI 35 condition
>> type has a particular fixed set of fields whereas much of the point of
>> 198 is that the set of properties can evolve and be freely chosen from.
>
> Wouldn't extra fields correspond to SRFI 35 compound conditions? Each
> set of extra fields corresponding to each other would become a new
> simple condition type packaged with the rest into a compound
> condition.

Using SRFI 35 you need to define a new condition type for each
combination of fields, right? 198 in its current form lets each object
freely mix any fields without defining a new type.

In principle we could define a separate SRFI 35 condition type for every
possible field, which would mean a 5-field SRFI 198 object becomes a
compound condition of 5 other conditions, but it gets quite unwieldy.
SRFI 198 properties also don't need to be declared ahead of time, so new
SRFI 35 condition types would have to be created on demand at runtime
and cached in a table.