thoughts

Core:
- store code in memory/graph db
	- graph-db or acid-state.
	- storage is controlled by it
	- what about SCM integration then? -> consistent source export
- store code as AST
	- typed lambda calculus + extensions
	- some kind of "simple" language
- get data from 
	- all packages in sandboxes
	- all packages in hackage
	-> reuse scion-browser code
	-> get source code from github/darcs site, etc?
	- local projects for which we have source code
- run code 
	- via interpreter?
	- via transformation to Haskell: existing haskell code we generate normal invocation our code is mapped to interpret

UI:
- use Web UI
	- threepennygui + other javascript library?
	- web services API implemented in Haskell + Javascript client
- search list
	- like hoogle search
	- nothing: list local projects
	- type in: find matching projects/types/functions
- project details
	- when clicking on a project
	- project info
	- project dependencies both ways
	- project contents
- folder details:
	- contents
- module details:
	- doc
	- contents
	- module dependencies both ways
- function details:
	- doc
	- code how?
	- add sugar
	- references
	- flags necessary
	- related instances
	- link to used functions/types definitions
	- if source code know, replace function invocation by code
- edit code
	- how?
	- structured or not?
- save
- export to Haskell project
	- and build