[1] 'A written sign, in the usual sense of the word, is a mark which remains, which is not exhausted in the present of its inscription, and which can give rise to an iteration both in the absence of, and beyond the presence of the empirically determined subject who, in a given context, has emitted or produced it' (Margins, p.317).
What does Derrida mean here? A written sign has the structural ability to signify in the absence of the addressee and independent of the (intention of the) sender. (According to Derrida, whatever applies to writing, applies to every sign and to the whole field of experience because it always takes place in chains of references.) In order to function as writing, that is, in order to be legible, any written communication must remain legible in spite of the absence of the addressee. A writing that would not be structurally legible - iterable - beyond the absence (death) of the addressee would not be writing. The same holds for the sender or the producer, and for the same reasons. 'For the written to be the written, it must continue to 'act' and to be legible, even if what is called 'the author of the writing' no longer answers for what he has written, for what he seems to have signed, whether he is provisionally absent, or if he is dead, or if in general he does not support, with his absolutely current and present intention or attention, the plenitude of his meaning, or that very thing which seems to be written 'in his name'' (Margins, p.316).

[2] A sign thus carries with it a force of breaking with its 'original' context. This force of breaking is not an accidental consequence but the very structure of the sign. (What would a sign be that could not be cited?) This applies to both the so-called real context - the presence of the writer and his intention - and the internal semiotic context: one can always lift a sign from the chain of signs in which it is embodied by inscribing it into other chains. This force of rupture is due to the spacing that constitutes the written sign, the spacing that separates it from other elements of the internal contextual chain.
In order for a sign to be able to signify, it needs to be repeatable in a different context. It is an essential property of all forms of appearance to be able to appear at a different place and a different time. Without repetition, without the return of the same in a different way, there can be no matter of meaning or significance (cf. IJsseling, p.18). Even reading and understanding can be regarded as a repetition, a resumption of what has been previously written or said. It is a resumption at a different moment than the moment at which it was written or said.
In order for a text to be read, the reader must extract it from the author's protection (decontextualization). The same holds true for speech: once uttered, it becomes available for interpretation, repetition, and reformulation even by those who were not actually present at the time the words were uttered. Contexts thus constantly change. At the same time, it requires that each signifier remains forever severed from any specific meaning (cf. Neel, p.115). This is what Derrida calls dissemination.

[3] It is a property of every sign that it can be cited. 'Everything begins in the folds of citation ... , the inside of the text will always have been outside of it, in what seems to be serving as the 'means' toward the 'work'. This 'reciprocal contamination of the work and the means' poisons the inside, the body proper of what was once called the 'work', just as it poisons the texts which are cited to appear and which one would have liked to keep safe from this violent expatriation, this uprooting abstraction that wrenches them out of the security of their original context ... To try to resist the removal of a textual member from its context is to want to remain protected against this writing poison. It is to want at all costs, to maintain the boundary line between the inside and the outside of a context' (Dissemination, p.316). Derrida calls the impossibility to prevent the removal of a signifier from one context to another, a general iterability (the word 'iterability' alludes to both the possibility of repetition and to the possibility of change and transformation or distortion). Every sign can break with every given context and engender infinitely new contexts. This does not suppose that a sign has meaning outside of a context, but on the contrary that there are only contexts (cf. Margins, p.320-1). By grafting familiar terms onto new contexts, those terms are indeed not separated from their 'original' meaning (since they would then have no meaning), but room is made for other meanings. By undermining the univocality of a term, concept or word in this way, room can be made for that, which cannot be conceptually captured.

[4] Derrida seems to deploy a shift from the primacy of the text to the primacy of the context as the complex of circumstances that affect the production and determination of a certain meaning: there is nothing outside context! Nevertheless, some comments should be made with regard to context. Or, as Derrida wonders: 'Are the prerequisites of a context ever absolutely determinable? ... Is there a rigorous and scientific concept of the context?' (Margins, p. 310). First, context itself is a text in the general sense (arche-writing in Derrida's terms). Context is not simply a pre-linguistic given; it is just as thoroughly textualized as the text itself. One cannot interpret a privileged text against some 'harder' reality, for that 'reality' is itself constituted by other texts.
Second, any given context is open to further description. There is no fundamental limit to what might be included in a given context. Science and philosophy aim to have complete control over the context of their field of study. But total context cannot be mastered, both in principle and in practice. Meaning is context-bound, but context is boundless. Derrida: 'No meaning can be determined out of context, but no context permits saturation. What I am referring to here is not richness of substance, semantic fertility, but rather structure, the structure of the remnant or of iteration' (Bloom et al, p.81). Each sign, each text has a context that determines the meaning. However, this context can never be completely isolated. Post-structuralism in general argues that context is in fact unable to arrest the fundamental mobility of signs for the reason that it harbors exactly the same principle of interminability within itself, the impossibility of its closure. Context can always be extended or augmented. Certainly there will be a cut-off point, which, for example, can be determined by the conventions followed by the community of interpreters, but this is always arbitrary.
Third, context is boundless in another sense. 'Any attempt to codify context can always be grafted onto the context it sought to describe yielding a new context that escapes the previous formulation' (Culler, p.124). Each sign and each text can be taken out of one context and put into another context. In discussing the quote, 'I have forgotten my umbrella', that appears in Nietzsche's Nachlass (unpublished writings), Derrida writes that thousand possibilities will always remain open to interpret this sentence (cf. Limited Inc., p.63). They remain open not because the reader can make the sentence mean anything at all, but because other specifications of context are always possible. Surely some contextual factors are less relevant than others; certain readings of Nietzsche's sentence are less probable, but the point Derrida makes is that any demarcation of contexts will always be arbitrary and debatable, and far from neutral when viewed from a political or social perspective.
Fourth, it cannot be taken for granted that the evidence that makes up 'context' is going to be any simpler or more legible than the text upon which such evidence is to operate. The notion of context frequently oversimplifies rather than enriches the discussion, since the opposition between a text and its context seems to presume that the context is given, that it determines the meaning of the text and that it is ready to act upon the text to order its uncertainties. But things are not that simple since it cannot be assumed that context is a presumption or a simple natural ground upon which to base interpretation. Like a text, context is not given, but produced; what belongs to a context is determined by interpretive strategies. Contexts are just as much in need of elucidation as texts; and the meaning of a context is determined by texts. The idea of context, posited as platform or foundation, invites us to step back from the uncertainties of a text. But once this step is taken, it is by no means clear why it may not be taken again; that is, context implies from its first moment a potential regression without brakes.