Joe's GUID dreams vs. a Shoebox
Fri Feb 14 09:04:45 CET 2003
Marked up documents with GUIDs can be neatly organized
and referenced if you are a little bit meticulous about creating
things. I tend to get started on a lot of things, but never
really polish them enough to be fully organized.
I have a more fluid idea about information. I collect it in
little bits: newspaper clippings about new restaurants, or
jazz or movies; websites or phrases discovered in my
spare time; current events that color my every day behavior,
phone calls about future events for my calendar.
When I need to communicate with someone else, I organize
it to convey the proper message. Structure and organization
make it easy for someone to digest new information that is
delivered en masse. But I collect it, think about it, and access
pieces of it for different reasons at different times. I could
never categorize it all neatly for all the reasons I might need
to access it in the future. I might publish an article using
factoids and then a week later rearrange them and produce
something else. What is essential to me is collecting and
rearranging the data, not structuring it once forevermore
and giving it an immutable name.
If I were a bank with 100,000 customers or a lawyer with
100,000 documents, I would want a highly organized data
base that employees can access interchangeably. SQL and
XML are all about that.
On my Personal computer I want a digital shoebox. I can throw
anything I come across into the shoebox. When I have spare
time I can rummage through it and look for patterns, find a
diversion to investigate or just do some serendipitous browsing.
It's hard to do that with structured data. I want the data to self-
organize based on my access patterns. This month may be
all about calling people and listening to jazz, next month may
be all about vacation planning. I want to collect the raw data
and investigate it 6 months from now when my interest finally
I think documents, files and organized containers are holdovers
from a past when data could only affordably exist in one format
and so the most generally useful access method became the norm.
The fact that source code is in files restricts the richness of
expressing software; documents with permanent GUIDs are like
stone tablets that are useful reference tools but not useful idea
formulation tools. Computers are fast enough now to adapt the
data to the use, rather than the other way around. The walls of
normalized SQL tables need to be broken down.
More information about the erlang-questions