The clause was the eleventh of nineteen, and it committed both parties to a joint commission that was never convened. This was discovered in 1974 by a graduate student comparing the two archived copies, which had until then been assumed identical. They were not. The English text obliged the parties to establish the commission *within two years*. The other text, equally authentic under the terms of the agreement, obliged them to establish it *when circumstances permit*. Neither delegation appears to have noticed, and there is no correspondence on the point in either archive. How this happened is reconstructible. The clause was inserted late, at the request of a minister who was not present at the final session, and translated overnight by a secretary working from a summary rather than the text. The summary was accurate; the translation of the summary was reasonable; the result was a divergence that nobody was in a position to catch, because catching it would have required someone to read both final versions side by side, and the two final versions were signed in different rooms. The historiographical interest is not in the mistake but in its consequences, which were nil. The commission was never established under either reading. Neither party ever invoked the clause, complained of its non-performance, or referred to it in any subsequent negotiation. For a hundred and ninety years an obligation existed, in two mutually inconsistent forms, and was without effect. This is more common than the diplomatic literature likes to admit. Treaties are frequently written to secure a signature rather than to be performed, and clauses of this kind serve a purpose at the moment of drafting: they allow a negotiator to return home with something to point at. Their subsequent non-performance is not a failure of the instrument. It is, arguably, the instrument working as intended, though nobody will say so in the preamble. The student's supervisor discouraged publication, on the grounds that the finding was trivial. She published anyway, in a minor journal, and the paper sat uncited for eleven years. It then became briefly important during an unrelated boundary dispute, in which counsel for one side argued that the discrepancy demonstrated a broader pattern of imprecision in the treaty and that its territorial provisions should therefore be read narrowly. The tribunal rejected the argument, but at some length, and the paragraph in which it did so is now the standard citation for the proposition that an error in one clause does not impeach the whole. The paper has since been cited two hundred and forty times, almost always for that proposition, and almost never for what it actually says about the eleventh clause. The original student, now retired, has remarked that this is the ordinary fate of archival work. You find something small and true, and it is eventually useful for something else entirely. She added that she had never expected the finding to matter, and that on the whole she preferred it that way, since a discovery that matters immediately is usually one that somebody was looking for.