I have been working on a site (www.polytechnic.wa.edu.au) based on
repoze.bfg running on app engine for the last few months.
(I don't do visual design I just make it run). It uses bfg-pages
(http://code.google.com/p/bfg-pages/) for zpt templating and appengine
datamodel traversal so
that it can reflect large amounts of the plone structure. All of the
custom entities used are designed in Enterprise Architect and the
python classes and interface definitions
are generated directly from the UML model.
Beaker is used for session management middleware with a custom plugin
to put sessions into memcache.
Formish is used for the few forms we use in appengine (mainly site
configuration and our own concept of actions and portlets for
Content is authored in Plone (running on EC2) which is then pushed
from plone to appengine.
Course data is extracted from a SQLServer database and pushed into
plone where it is annotated and then published to appengine
All content is indexed on solr (also running on ec2) for searching
from appengine (appengine doesn't have any text search capability
worth mentioning) .
We had a few teething problems in the last week where our production
instance's performance was being affected by things happening inside
Which turns out be a known problem (for google engineering that is ;-)
In the last 24 hours we moved all of the data to a new instance and
all seems well so far (fingers crossed).
Repoze-dev mailing list