Halliday & Matthiessen (1999: 12):
There is no separate component of “pragmatics” within our interpretive frame. Since it emerged as a distinct field of scholarly activity, pragmatics has by and large been associated with two aspects of language.
On the one hand, it has dealt with those aspects of the meaning of a text which depend on specific instances — particulars of the situation and of the interactants, and inferences drawn from these. But just as, in grammatics, we do don't distinguish between the grammar of the system and the grammar of the instance — a systemic theory is a theory of both, and necessarily (therefore) of the relationship between them — so in semantics we would not want to separate the system from its instantiation in text. In this respect, pragmatics appears as another name for the semantics of instances.
And on the other hand, pragmatics has served as an alternative term for the interpersonal and textual domains of semantics. Here the distinction that is being labelled is one of metafunction, not of instantiation; but it seems undesirable to obscure the relationship between ideational meaning on the one hand, and interpersonal and textual meaning on the other hand by locating them in different disciplines.