Brian Aker wrote:
Hi!
On Aug 20, 2008, at 2:27 PM, Baron Schwartz wrote:
If you use the storage engine to also commit DDL changes, commit
handling of
DDL is simplified. With separate DDL persistence, you need a kind of
2PC. An
Drizzle uses, well will use completely when we are done, "discovery".
The engines are asked if they have a table, not any global cache/system.
We will have a default method (which will be something a bit more then
what we have today with FRM), but I expect most serious engines to
handle this data on their own.
Which part of the system will do logical schema mapping, ie map a
physical table into a logical schema name dot table name?
If you allow complete freedom, then two engines may insist on serving
the same table names...
And, obviously cross-engine FK will be difficult when the engines are in
complete control of their table definitions :)
Another thing I would consider is to let all communication between
server and engines go through negotiated handles, and use methods to
query the properties of the objects underlying the handles. This way,
the server and engine will not share any data structures (except for the
handles), and are free to create new versions of internal data
structures without demanding new versions of other components.
But row and field data should probably be communicated in a well-defined
format, for efficiency reasons.
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
