[Mailman3-dev] Flexible data storage

[Barry Warsaw]

On Mon, 2004-03-15 at 15:50, Dale Newfield wrote:

> The questions of "who defines what nouns (tables) exist," "who defines how
> those nouns (tables) translate into mailing lists," and "who manages the
> nouns (rows in tables)" are critical here.  If MM is to function, it
> clearly needs to expect to be able to determine some nouns and manage
> them.  

This is the primary thing I want to come out of the sprint with: an
interface, defined in Pythonic/Zopeish/PyProtocolli nomenclature, of the
types of data that Mailman will want to read and write.  This will be
the primary communications mechanism for those writing the db backends
-- which will have to map those interfaces to whatever storage formats
are most appropriate -- and the Mailman source code, which will not care
about the implementation of the backends.

Of course, there are tricky bits at the edges such as, how do you define
which storages Mailman actually uses?  How do you create them?  How do
you cleanly close them?  How do you define transaction boundaries?  How
do you manage transactions across multiple different data storages?

I have some stuff about this in the MM3 source, which I will soon be
checking in.  I'm not entirely happy with it...

> This would
> enable the "student" table to exist within the app, the "subscriber" table
> to exist within MM3 and linkages between them.  I've attempted to write
> about this in a general way several times before and feel I've never
> succeeded well.  This is why I'm looking forward to the sprint.

Me too.  It's hard stuff.

-Barry