I have built a *simple* object store that runs on the local filesystem.
Object serialization (marshaling) and restoring details are pseudo
automagical, serialized objects saved as binary object strings in text-
files. A label is generated for indexing and metainformation.
You can save, search for words in label fields, and restore objects.
I implemented it by making all storable objects inherit from the storable-
superclass. Least effort -principle. Far from a database.
Serialized objects are all stored in the same directory. After a few more
thousand files I suppose directory size limits performance. And what
about the relationships between the objects?
I want to learn more.
What laws dictate performance? How should filesystem<->objects -mapping
work? In effect, what kind of hierarchy and naming should, in your
opinion, be used?
I realize its not that simple but where should I look for answers after I
find the questions?