Magma + Seaside

[Cees De Groot]

On 10/27/05, Chris Muller <chris at funkyobjects.org> wrote:
> This will be fun.  There is no one with your experience using M+S, and I'll be
> glad to learn more about Seaside, thanks!
>
Why do I always need to walk along the bleeding edge?

Anyway, I've been busy yesterday night moving some code from
VisualWorks to Squeak, so I only had a short while to look at M+S.

The only/biggest issue is probably session handling. OmniBase allows
multiple parallel transactions on a single database connection, so in
every Seaside request handler, you simple grab the application-global
connection, create a new transaction, and commit it at the end of the
request handler (unless there's an error, then it does a rollback - of
course, one could do it just as easily the other way: abort by default
and commit on a special notification. The idea is that the request
handling code is wrapped in a transaction).

With Magma, it looks like the best approach is to use a connection
pool of Magma sessions. I have a simple, but servicable generic
connection pool class (one of the things I moved over from TIO), and
that's probably going to be my first attempt.

The idea is to include Mewa as well in the experiment, as well as the
Timetravel code I ported over yesterday (see SqueakPeople). The end
result should be a simple environment where you can immediately start
writing a domain model and seaside compoments without having to worry
about persistence - I want to rid myself of having to be concerned
about that level, doing demos based on image persistence, etcetera,
for once and for all :)