1- A conventional system of representation
2- A method of externalising the given conventional system of representation i.e. Through speech or writing, sign language, covert communication through general facial expressions.
It does depends on what you mean by "language". It's a very broad term, but i assume you mean a conventional language that you speak.
Noam Chomsky sees 'Language' as a completely different abstract entity to what we call 'Language-s'. Although i think with the DIP (double interface property) he would also agree on the grounds that there are two facets to 'The Sign' - these are the Logical Form (LF) and the Phonological Form (PF). Chomsky says that these two form what is known as 'The Sign', this in short is merely 'speech' or an externalised speech.
But if you mean 'Language' the way that Chomsky means..this refers to a 'Language' of the mind.. this is a language you can not externalise but serves as the purpose to generate 'realizations' (to make real) of language.. Chomsky sees the internal language as completely isolated, but then he says speech is instantiations (instances) of the internal language. This is the weak point of his theory. When we speak, he says we are 'realizing' the internal language.. this doesn't make a lot of sense though when this internal language is meant to be unexternalisable.
Noel Burton-Roberts proposed 'The Representational Hypothesis' This goes on the basis that we do have an internal language, but speech/writing are mere conventional representations for the internal language. This allows us to say that speech doesn't contain meaning or semantics, which as absurd as it may sound, is actually true.. it's not the speech that carries meaning, it's the listener that infers meaning.
With Chomsky's theory, we'd have to say speech carries meaning, it carries semantics and all the components of the internal mind, and this is all conveyed through the air when we speak. The idea itself is completely flawed as this fact is untagible. 'Meaning' can't travel through the air! It's an abstract entity!
With NBR's representational hypothesis he says that we have a phonological system which contains all the instructions to give to the speaker in order to tell them how to represent the 'i-language', this again moves away from a very internal syntactic component feeding information directly in to externlisable language.
Chomsky believes that we have an innate internal component for languages, called the HLF (Human Language Faculty). This is set to accomodate for the acquisition of languages - chinese, english.
Internal 'Language' (I-language) i'd say is a very different thing altogether from languages themselves. So beware of the term 'Language'...