[Seaside] Meta models

[Cees de Groot]

On Sun, 2003-04-13 at 03:09, Todd Blanchard wrote:
> EOModels are great meta models [...]

I think - well, I hope ;-), that Derek was referring to something more
lightweight. In Sagan there's currently a bit of lightweight
frameworking that helps in giving classes more descriptive attributes,
with automatic timetravel over them, etcetera. 

Personally, I think we have a reasonable modelling tool - the browser.
The only thing I really miss is some simple metadata like the domains of
attributes (either physical - instance vars - or virtual - getter/setter
pairs), support to generate user interfaces, stuff like that. That's
what I'm trying to add in Sagan. 

At the moment, the basic framework is ready, but two things are missing:
- there's just 'Attribute', some sort of type hierarchy must be built up
around it. This is not so much a matter of a lot of work, but rather
it's a matter of choosing the right approach;
- the browser needs to be extended to take these attributes in
consideration e.g. when looking for implementors of names that match the
method signatures the DNU handler deals with. 

In my 'daytime' project, which uses an older and therefore much messier
version of this, we are now on speed with automatically generated
components around our 'business objects' based on Attribute information,
including security information (certain roles may or may not see certain
attributes, etcetera). No automatic navigation yet, but hooking up views
for navigation is typically a one-liner.  So, the whole 'browse' view of
our object database comes largely from the metadata. More task-oriented
things, like our on-line shop, are of course still hand-crafted (worse,
they still live in .ssp land...). 

For Sagan, in order to get a clean and minimal model, my idea was to
move up to the domain model first (modelling users, rights/capabilities,
various 'data' sitting around like folders, files, etcetera) and just
refactor Attribute until it could support the domain model. I think that
if anyone else wants to use what's there and extend/complete it, that's
still the most sensible approach. The whole framework so far has been
1:1 copied from my VisualWorks daytime work which was 'grown' by
refactoring, and I have carefully tried to extract only the clean bits
so what is currently there presents 'real world knowledge' instead of a
'wannabe framework'. I stopped at Attribute because what we have in VW
currently is a mess :-)




-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://lnx-12.ams-2.theinternetone.net/pipermail/seaside/attachments/20030413/a1bdc68c/attachment.bin