This is an interesting approach for primary key obfuscation.
I recently stumbled upon something that I feel might be even more powerful - Hashing complex types into 256-bit keys. E.g. If you had some type representing the composite keys of a Customer in your system (email address, phone number, etc.), you could serialize this instance, run it through SHA256, and as long as the same process is used for lookups, you can get everything back out as expected.
Essentially, you can compress your entire scope of composite key data into a single 256 bit value. Just like with GUID keys, this can be pre-computed on each client (whereas autoincrement cannot). This approach is very clever IMO because it can be used directly on top of any universal byte[]/byte[] key-value store. Your keys are all 256 bit values corresponding to the SHA256 of a serialized complex key instance. The type information can be encoded into the key itself (e.g. hash the fully-qualified type name as well).
I recently stumbled upon something that I feel might be even more powerful - Hashing complex types into 256-bit keys. E.g. If you had some type representing the composite keys of a Customer in your system (email address, phone number, etc.), you could serialize this instance, run it through SHA256, and as long as the same process is used for lookups, you can get everything back out as expected.
Essentially, you can compress your entire scope of composite key data into a single 256 bit value. Just like with GUID keys, this can be pre-computed on each client (whereas autoincrement cannot). This approach is very clever IMO because it can be used directly on top of any universal byte[]/byte[] key-value store. Your keys are all 256 bit values corresponding to the SHA256 of a serialized complex key instance. The type information can be encoded into the key itself (e.g. hash the fully-qualified type name as well).