[<<][pool][>>][..]
Fri Jul 14 18:24:40 CEST 2006

databases

more computers, more problems to glue it all together. i seem to be
converging to the following tools:

* code in darcs, including scripts and config
* compiling config files

my current problem is managing mac addresses for all sorts of
things. i need some database and maybe it's time now to start using
the filesystem as a database with lots of small files, following
reiser's ideas a bit.

this maps easily to scheme code, and i already use it in CAT, which
has 3 directories: code (ro), state (persistent rw) and current
(working memory).

one thing to find out is how to use emacs to edit/navigate a tree of
small files.

hmm... reading http://www.namesys.com/whitepaper.html it seems to be
about a lot more than making small files work fast :)

3 classes of organization
- hierarchical
- relational
- semantical (keywords)

anyways, small files solve my particular problem as they serve as a
unifying syntax for variables in scripts. making them map better to
the language (scheme or CAT) makes things easier to use, since parsing
and ad-hoc syntax can be largely eliminated.

i'm not sure how to implement invertable relations though. in
databases, this is implemented by explicitly storing both directions
in tables. for the DNS problem, i have projective maps from names to
unique-name/ip/mac, where i should be able to relate all 3 components
to each other.




[Reply][About]
[<<][pool][>>][..]