I was looking at Isidore's Etymologiae today and had an insight into the medieval practice of etymologia. I've noted before that it's a mistake to overhistoricize what the medievals were doing with their etymologies. For us, etymology is a historical reconstruction, but while the medievals thought there was some kind of rough historical connection, the medievals aren't trying to reconstruct the history. For them, etymologia is not a historical origin, per se, but an imperfect definition -- it is a definition-like thing, not a history-like thing, that falls short of the full conditions for a good definition. In particular, if you are giving the etymologia of a word, you are defining it in terms of similar words.
Now, this is to us an odd thing to take seriously; why would there be any particular importance in using similar words? If we were doing it in English, it would usually be a game. Nonetheless, you still can find cases here and there where it's obviously relevant. I've noted before it's actually essential to understanding 'parameter' that people in colloquial English take parameters to establish perimeters, even though the only reason for this is that 'parameter' and 'perimeter' are similar words. English is blocky language, so it's even easier to see in common phrases than in particular words: the common misuse of the original technical term 'begging the question' is inevitable, and even people who insist on keeping the technical meaning will often, if you ask what the phrase means, try to fit the 'begging' in somehow, even though it's only there due to an oddly strong and overliteral translation of the Latin petitio.
But, of course, the medievals were not thinking in terms of English but in terms of Latin and sometimes Greek. And, structurally, what is Latin like? It consists of roots, prefixes, suffixes, case endings, and the like, and shortened forms are actually fairly common. So what St. Isidore is doing is just rationally extrapolating this to the limit, and proceeding on the assumption that every non-basic Latin word consists of further roots. All Latin words break up into little bits anyway; lots of those little bits obviously contribute to the meaning; it at least raises the question what the other little bits might be doing.
So, for instance, Isidore's etymology for gladius, sword, is gulam dividere, splitting the throat. What is his reasoning? He's not merely playing a word game. As he sees it, gladius breaks up into smaller elements:
The component g(*)la is shared with gula; the component di is shared with dividere -- and, indeed, it is found in lots of Latin words that have something to do with dividing. So we get gulam dividere; and since it makes sense to think of a sword as a throat-splitter, he takes that we have here come up with a plausible candidate for more basic words using the same components that capture the meaning of gladius at least roughly. (And Isidore is never really dogmatic about his etymologies, often willing to propose alternatives, although some of his proposals are so catchy that once he proposes them they become how people think of the original words anyway.)
A bigger stretch is spes, which he explains as pes progrediendi. Spes and pes obviously share a component; so to make sense of how this component in 'foot' can apply to 'hope' as well, we need to ask, "What hope-like things does a foot do?" And Isidore's answer is the obvious one: it moves forward.