> From: Lassi Kortela <xxxxxx@lassi.io>
> Date: Friday, August 14, 2020 10:57 AM
>
>>>     category
>>>     class-code
>>>     class-title
>>>     subclass-code
>>>     Subclass-title
>>>
>>> All these things are determined by the SQLSTATE, so I don't see much
>>> reason for lugging them around in status objects.
>>
>> I say leave it at the discretion of the library developer, and for now
>> many of them are used in one SRFI example.  Lugging them around may save
>> the end user a trip to a file or book.
>
> OK, let's put them in the registry but make them strictly optional. A
> particularly conscientious database driver programmer can fill them in.

In my mind, one purpose of registries starting with the basic one at
the end of SRFI 198 (see Standard facility conventions in
https://htmlpreview.github.io/?https://raw.githubusercontent.com/hga/srfi-198/master/srfi-198.html)
is to define optional keys that are not optional if you so choose to
report a datum value that is defined to have a specific key.  So it
defines (for now, very subject to change or moving to Schemeregistry)
'source-filename, 'line-number, and 'column-number.

> Well, class-code and subclass-code are simple parts of sqlstate so
> maybe they shouldn't be there? We can decide once we're actually
> writing drivers, SRFI 198 doesn't need to talk about SQL stuff.

Per my above paragraph, they'll perhaps be there someday, but as you
say, only after we reify our discussions of SQLSTATE into real drivers.
Which I'm not far from doing, obviously gated on finalizing SRFI 198.

> [ Mnemonic is out. ]

- Harold