Terry Hancock, a frequent poster on the law-oriented cc-licenses list, is working on an interesting metadata library called Palimpsest:

[W]hich has a mnemonic association with what the program does, and does have a clever backronym for those who want one:
P ython
A ttribution &
L icensing
I nformation
M etadata
P rocessor, with
S ystematic
E xtensibility for
S undry
T ypes

Terry's goals for the project:

• Read/write support of Adobe XMP embedded metadata
• Read/write support of native “named field” data
• Read/write support of comments
• Read/write support of visible text labelling for formats that need it
• General adaptation to the 15 Dublin Core named fields for all data
• Discovery of attribution and licensing data in comments and annotations, if not available elsewhere
• License-aware processing (expansion of common abbreviations of terms, etc)
• Open-ended pluggable support for virtually any multimedia datatype
• Highly portable, so that it can be used on clients or servers on any operating system
• Dead-simple, so people will actually want to use it

I'm glad to see Terry tackling this project. It'll be hard to get the abstractions right, but valuable if it works.

I love the project logo:

Not because it is a particularly great logo, but because it's the first logo I've seen that could be mistaken for a captcha. Intentional or not, bound to be independently invented many times, and perhaps copied by me at least once.