Preventing metadata hash collisions

[psiemens]

The current NFT metadata serialization scheme is prone to duplicate encodings (and therefore duplicate hashes) when used with arbitrary-length values (e.g. strings).

For example:

"foo".concat("bar) == "foob".concat("ar")

In this case inputs are distinct but the concatenation is the same.

[psiemens]

To solve this, I am likely going to add a constant-length size prefix to the following types:

• String
• Int
• UInt

The remaining fields types (e.g. UInt64, etc) have a constant-length encoding.