23.11.2018 23:12, Dimitry Sibiryakov wrote:
can store sequence number elsewhere (and it did until recently), but
it's more error-prone to store it inside the database itself. Header
page is an obvious choice. RDB$DATABASE is another option, but it's
also an ODS change.
Isn't a sequence even more obvious choice?..
It could be, but new system generator is also an ODS change. IIRC, we
don't have a safety gap between system and user generator IDs.
BTW, database encryption already has access to a database ID.
Replication database ID will be the same or another?
I probably missed that. Do you speak about dbb_id (filesystem level ID)
or something more persistent?
I introduced a general GUID that potentially could be used for multiple
purposes. In replication it helps to provide global uniqueness for
transaction IDs, i.e. it should persist when database is copied and must
be reset when database is restored via gbak. Do we already have
something like this?
Also isn't function of recently invented SCN the same as your journal
number? Don't they match?..
No, they are different things. The sequence I was talking about is just
to provide contiguous numbering of journal files.
Dmitry
Firebird-Devel mailing list, web interface at
https://lists.sourceforge.net/lists/listinfo/firebird-devel
