An interesting phenomenon of language variation is lexical reversal, where a term that normally points in a specific temporal direction is flipped. Hence people will occasionally use ancestor, which points backwards in time, where descendant, which points forward, would be standard. Arnold Zwicky has commented on this phenomenon a number of times at Language Log, providing examples such as this one:

Fred Willard, director of the Lost Colony center, said some colonists may have migrated inland to what are now East Lake, Chocowinity and Gum Neck. Researchers plan to use cheek swabs taken from possible ancestors to test the paternal and maternal DNA lines.

Zwicky suggests that this switch represents "a desire for double-sided lexical items," that is, for items that can point indifferently to the past or the future. A classic example of this is the Hindi word ?? (kal), which means either yesterday or tomorrow depending on the context. If there is such a "desire," though, it has to be fairly weak, since such terms are relatively uncommon, and there seems to be a strong countervailing pressure to give double-sided terms directionality. As a case in point, consider the word namesake. Although different dictionaries give somewhat inconsistent definitions of this word, it's safe to say that traditionally, it has long been standard to use namesake to refer either to the person named after another (the successor) or to the person after whom another was named (the predecessor). Paul Martin, in Guide to business style and usage (2002) is quite confident, however, that only one of these is correct:

namesake A person is the namesake of the person he is named after. Although the person whose name is taken by another is sometimes also called the namesake, avoid that usage. (p. 163)

Like many usage authors, Martin provides no support to back up his opinion, so it's hard to tell where he derived the notion. Namesake is not a traditional subject of usage complaints. The Merriam-Webster Dictionary of English Usage, for example, has no entry on it. Apart from Martin, I have found no other books that censure the second use. If he's right, however, the predecessor use of namesake would count as another lexical reversal. One place from which he might have derived the notion is the American Heritage Dictionary of the English Language, which does indeed define namesake as "one that is named after another." Yet if you look at other dictionaries, not to mention actual usage, it becomes clear that the American Heritage definition is inadequate. Merriam-Webster has the following definition: "one that has the same name as another; especially: one who is named after another or for whom another is named" That definition seems to accurately capture the range of meanings in actual use. I audited the first 200 hits on Google for the search {namesake -definition} and classified results into one of five categories: predecessor (i.e. namesake refers to the source of the name, successor (it refers to the one who receives the name), both (the same work uses namesake as both inspiration and receiver in different places), same-name (namesake used for two people sharing a name without implication that one was the inspiration for the other), and other, for instances where the intended meaning can't be determined from context. Repeated references to the same book, article, etc. were excluded. Thus although a large number of hits referred to the Lahiri book The Namesake or the movie based on it, I only counted it once. For pages that aggregated different usages that were not written as a single piece, each example was counted separately. Here are the results (percentages don't include the "other" category): Predecessor: 26 (34%) Successor: 32 (42%) Both: 2 (3%) Same name: 16 (21%) Other: 16 Clearly, the Merriam-Webster definition accurately captures real usage, and if this small sample is at all representative of overall usage, it's misleading to assume that there's a marked preference for the successor meaning. The historical usage also fails to help Martin's claim. The OED entry defines namesake simply as "A person or thing having the same name as another" and does not specify the tendency to use the word when there's a specific relationship between the source and the receiver of the name. In this meaning, the earliest entry provided is from 1646: Sir T. Browne Pseud. Ep. 170 Nor [does] the Dog~fish at sea much more make out the Dog of the land, then that his cognominal or name-sake in the heavens. And some other early citations fit with both successor and predecessor meanings respectively: 1657 J Watts Vindic. Church Eng. 89, I shall here dehort you from being of Iohn and Iames (though you are the name-sake of the one). 1712 Addison Spect. No 482 ¶2 Another . . subscribes herself Xantippe, and tells me, that she follows the Example of her Name-sake. There is also a somewhat earlier usage in the phrase "for one's name('s) sake" or "for name sake", but in this sense, there is not the notion of a shared name: 1526 TYNDALE Acts ix. 16, I wyll shewe hym howe grett thynges he must suffer for my names sake. Assuming this phrase is the direct source of namesake, it's a little unclear how you get from here to Browne's usage (the OED tentatively proposes that two individuals are connected for the sake of (i.e., by means of) their name), but the word's later development suggests that the phrase remained in the back of speakers' minds, as the word acquired the sense of being named in honor of someone else. As that meaning was added, some speakers also seem to have decided that the term had to be directional, for example, referring only to successors. But as the earlier meanings do not specify a temporal direction, the use of namesake to refer to a predecessor is not, in fact, an instance of lexical reversal, as in the ancestor/descendant confusion. Rather it's a case of narrowing, and it reinforces the notion that double-sided terms are somewhat unstable, and that there's a tendency to want to make them more specific.