> On 6 Dec 2021, at 6:05 pm, John Levine <jo...@iecc.com> wrote:
>
> This is the IETF's forwarding addresses for works that are or were
> (hence the large number) in progress. Each entry has as many targets
> as the draft has authors so it's rarely more than 2 or 3.
You will not face any meaningful limits with this. Use of CDB will
be more efficient in memory cost, because unlike Berkeley DB the
database is read-only and does not allocate a page pool. Just
mmaps the (shared) file.
I'd use CDB for this. I think the inputs will not change frequently
enough or be anywhere near sufficiently many to make the CDB map
creation time to be something to worry about.
CDB has a very stable disk format and API, I trust it more than
either Berkeley DB or LMDB for mostly static data.
--
Viktor.
