An automatic cataloging system.
What if we created a catalog key type who's data could be
appended to by anyone as long as the appendix was a single
line that conforms to a specific syntax, is not already
contained in the catalog, and does not exceed a well known
reasonable length (I propose 256 characters). This would
not eliminate malicious abuse of the catalog files, but it
would help reduce the severity.
When a KSK is inserted, the client submitting the insert would
also update the appropriate catalog documents. There would
be two types of catalog documents, one for lists of categories
and one for lists of documents. If a user inserts
mp3/metallica/EnterSandman.mp3, the client attempts to update
the mp3/metallica/ document catalog to include
EnterSandman.mp3. If the catalog doesn't exist, one is
created and then the mp3/ directory catalog is updated to
include metallica/. If there is no mp3/ directory catalog,
that is created and so on. The root directory catalog would
be some well known file like /.
To maintain the catalogs, whenever a data request passes
through a node and the request fails, a quick check is made
to see if the key is in any locally stored catalogs. If so,
it's removed from the catalog. Catalog entries can only be
removed by the node the catalog is stored on. This prevents
anyone from going around deleting catalog files and still alows
for bogus or old entries to be removed.
l8r
Aaron
_______________________________________________
Freenet-dev mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/mailman/listinfo/freenet-dev
