Sat Jan 9 09:03:21 CET 2010
I don't want to kill the current ramblings implementation. I wonder
if a full rewrite following a different model isn't a better approach.
The graph n@ n! code is embedded quite deeply.
The problems ramblings.ss actually solves:
- article parsing
- article views (html, rendered .tex)
- list of articles (a ramblings.txt file's default order)
- data management (memoization or dependency management)
I also miss a decent emacs interface, and a definition of a single
aggregated context object (i.e. text + embedded non-text objects), and
a coupling to version control. Also, authentication would be nice.
I need an object model and a concrete data store. The store should be
simple to update, and hopefully human-readable. Large text files are
nice, but not quite managable.
So, what is an article? Essentially an object with different views.
Wait.. Can the current representation be converted to
single-assignment form? I like the memoization & cache approach, but
a static name structure would be simpler.
Also, indexing needs to be separated from construction. Maybe that's
the place to start first? Add one layer of indirection there?
Yes, this is really to complicated. The cross-linked objects model
seems to work in practice. However, I can't prevent myself from
thinking it's a ball of mud. You can't cleanly separate out a part,
but mayble that's the point?
This is really for later.. I need a clearly motivated goal to tackle