Re: Perform streaming logical transactions by background workers and parallel apply

On Tue, Jan 24, 2023 at 9:13 AM Peter Smith <smithpb2250(at)gmail(dot)com> wrote:
>
> 1.
>
> IIUC the GUC name was made generic 'logical_replication_mode' so that
> multiple developer GUCs are not needed later.
>
> But IMO those current option values (buffered/immediate) for that GUC
> are maybe a bit too generic. Perhaps in future, we might want more
> granular control than that allows. e.g. I can imagine there might be
> multiple different meanings for what "buffered" means. If there is any
> chance of the generic values being problematic later then maybe they
> should be made more specific up-front.
>
> e.g. maybe like:
> logical_replication_mode = buffered_decoding
> logical_replication_mode = immediate_decoding
>

For now, it seems the meaning of buffered/immediate suits our
debugging and test needs for publisher/subscriber. This is somewhat
explained in Shveta's email [1]. I also think in the future this
parameter could be extended for a different purpose but maybe it would
be better to invent some new values at that time as things would be
more clear. We could do what you are suggesting or in fact even use
different values for publisher and subscriber but not really sure
whether that will simplify the usage.

[1] - https://www.postgresql.org/message-id/CAJpy0uDzddK_ZUsB2qBJUbW_ZODYGoUHTaS5pVcYE2tzATCVXQ%40mail.gmail.com

--
With Regards,
Amit Kapila.