Tim Soderstrom wrote:
On Apr 17, 2009, at 8:56 AM, Baron Schwartz wrote:
Please, please, no! I_S != storage engine
The actual interface supplied to an I_S would not be the SE
interface. It
would be built on to the same execution path though (unlike the current
solution which has its own execution path).
Why not *actually make it a set of views* instead of "pretending" it's
a set of views.
I know of one other open-source database whose initials stand for
PostgreSQL that (I've been told) implements I_S as a set of views.
Nothing special about them -- just normal views, as I understand it.
(This is easy to verify or disprove, but I'm lazy).
I think it does, but those views are actually pulling data from
PosgreSQL's own informatin/performance schema thinger. No idea how they
implement that, though :)
This is also how we did it in Clustra/HADB. Basic schema information was
stored in base tables. DDL commands and name resolution used internal
interfaces to these tables. User-level information retrieval was through
the system-defined views (all tables have no public grants). The views
were not standard compliant INFORMATION_SCHEMA, though.
Even the Oracle-like DUAL table was implemented as a view...
Thanks,
Roy
_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help : https://help.launchpad.net/ListHelp
_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help : https://help.launchpad.net/ListHelp
