|
|
I would like to discuss alternatives for the generation of unique
identifiers, in the context of URLs, database primary keys/no-sql
keys, system interoperation, etc., etc.
I used GUIDs in the past, which for simplicity I stored them in its
string representation instead of using the byte/long representation.
Do we have something other than UUID (16byte/36chars) to create identifiers?
Does exist a shorter (ie. 8byte) UUID version?
I know its uniqueness comes from the extra bits, but maybe a shorter
display string can be used (as git commits does, but with 20 bytes) or
directly less bytes like MongoDB's OID (12 byte).
I'm thinking about unique identifiers in the context of different
systems (and different languages), where datatype limitations can
exist. E.g. Java's long is 2^64 -1, 8byte, SQLite can only store 8byte
integers. So the most standard generator, least common denominator
type would be the best approach.
By now, the "least worst" solution I found is to keep using the 36
char string representation for persistence/serialization (JSON). But
I'm sure there must be something else much more clever.
What do you use for IDs?
Regards,
Esteban.
Esteban A. Maringolo
|
Locked
