Popular posts

Actually, both are correct depending on certain situations.

Let's look at both perspectives.

One of the primary reasons for considering ID is due to it's immutable nature that it does not change overtime. Having said that, it's not necessary that you must use ID all the time, instead look for the immutable nature of a property which does not change overtime and consider it as Identifier of the object. This use case usually arises in static data models (for ex: cities in your case) where all you have is City Name as the property of the object, here you can store city names as Identifiers, for ex: New York can be stored as NEW_YORK.

• Pros: There wouldn't be a need of additional lookup from cityID to cityName every time you need to fetch city name for sending the response. These properties can be directly used in the analytics triggered from the client side.
• Cons: As you stated, any update of City name would result in updating existing collections, more the no. of places cityName is being referred, more the headache. Secondly, this approach might restrict extending the object with more properties in future where pointing to City name as the unique ID for that object would look weird.

Second approach, setting ID as the primary unique identifier of the object, is a standard and safe approach anytime with a compromise of being okay with overkill in the situations of static data models as mentioned above.

• Pros: Safe for extending the data object any point in time. Updates to any other properties of the object would not affect any other collections (assuming the updating property is not explicitly part of some other object as well otherwise you'd end up updating it everywhere).
• Cons: There exists an additional lookup every time you want the primary property of the object, specially in static data models.

So in a nutshell, look for the immutable nature of the property and the possible of extending that data model in future; based on these two factors, you can go for the appropriate approach.

P.S. If population of cities is the concern, you can always introduce Application level caching(redis or memcached) or better, Network level caching(max-age: <expiration-seconds> header in the http response - do this only if you're sure that your http response wouldn't change before given expiration duration)

If you need any more help, let me know. Cheers!