In what ways would it be different from the implicit assumptions we have concerning language, since we use sound?
Verbal language has many components. These components can be arranged in many ways in order to put some sense on the environment. For example, if you are car shopping, you need different names for the car, the door of the car, the price of the car, the suspension, comfort, price, colour, the type of fuel it uses, efficiency, age, price, capacity. . . .
A naieve scent-based language replaces each name with a scent. A conversation consists of two individuals exchanging short puffs of scent at each other.
However scent based language does not need so many names because it can interact directly with the environment. Rather than saying "Car door is good" you spray the door with the scent for "good". To say "Car door is very good" you spray the same scent with greater potency.
The speakers have very sensitive noses/tongues/antennae and can at shot range distinguish between different scents sprayed on different parts of the car. TO communicate at long range you need an extra large cloud of scent. "Shouting" like this is considered rude as it disrupts all nearby more delicate scents.
You have a smaller vocabulary but a greater number of ways to use each word. You could probably get by in a conversation with 5 or 10 words which carry no primary meaning alone, and only gather meaning when sprayed on two separate pieces of the environment. The thrust is either "these are similar" or "there are different". With enough reflection you can realise this is not too far off how humans communicate. We mostly just draw parallels between different things.
Consider how an ant colony self-regulated by pheremones.
Abstract concepts are harder to convey. To overcome this fact you have to get to the root of why abstract concepts are necessary in human conversation, when any outcome of the conversation can only be concrete and not abstract.
Spraying the scent for "bad" on someone has extra impact due to the physical violation required.
In spoken conversation it is considered rude to suddenly change the subject. The polite thing is to follow on directly from the other person's last sentence. Or at least seem to. Most 'etiquette' is a way of circumnavigating this fact.
By comparison, in a scent-converation every word lingers in the air. "Car door is good" remains visible long after it is spoken. At any point someone can go back to the door and spray on another scent to resume that thread of conversation. Thus scent-based conversations would have a degree of nonlinearity about them not present in verbal conversations.
Exercise: How can remote communication function for this language. A smell-making device is a must. But how does it capture the different locations of scents when the speakers are in different environments?