tea.mathoverflow.net - Discussion Feed (Are the "norms and metrics as parts of some whole" questions appropriate?)

salvo.tringali comments on "Are the "norms and metrics as parts of some whole" questions appropriate?" (19141)

salvo.tringali — Sun, 27 May 2012 03:59:49 -0700

@Andreas. Yes, surely! The framework should not be too general if the goal is to make it *useful*. Here, *useful* means, for instance, that typical theorems or constructions commonly encountered in this or that particular context - and I have listed normed groups, valuated rings, etc - can be proved or accomplished once and for all, without any need to replicate them in every each case as if they were totally independent from each other (that's why I made explicit reference to products and coproducts in the OP). From a more philosophical outlook, a "good unifying framework" should suggest new directions of research and provide insights for a better understanding of the ontology of metrics and norms (in the same spirit of Lawvere's work on metrics and metric spaces), e.g., by smoothing out, from the natural perspective of categories, a certain degree of arbitrariness which *happens* to be inherent, at least to my eyes, to the classical approach to the foundations of the theory of metric spaces (I've already alluded to this in reply to a comment by Pietro Majer), to the extent of separating, as I tried to say somewhere else, what is essential from what is accidental (in our standard vision).

Andreas Blass comments on "Are the "norms and metrics as parts of some whole" questions appropriate?" (19138)

Andreas Blass — Sat, 26 May 2012 16:02:33 -0700

Presumably, the intention was to ask about attempts to fit metric spaces and normed structures (of various sorts) into a single framework *without* making that framework extremely broad. (If extreme breadth were permitted, one could propose category theory or Bourbaki's treatment of general "structures" as examples of such frameworks, but I'm reasonably sure that is not the intent of the question.) It seems to me that a reasonable common framework, for any two notions, should not only incorporate those notions but should also allow proofs of some general results that specialize to something interesting for each of the two notions. I don't see much hope for that in the present situation. The difficulty is that, as far as I know, norms are used only in the presence of some algebraic structure (Salvo has listed numerous examples), and their interesting properties (apart from producing metrics) involve algebra, whereas metrics arise in far more general situations, where there might be no relevant algebraic structure. So if someone asked me what is the "most natural" general concept that subsumes both metric spaces and normed structures, I'd be inclined to answer "metric spaces" --- undoubtedly not what Salvo wants.

salvo.tringali comments on "Are the "norms and metrics as parts of some whole" questions appropriate?" (19137)

salvo.tringali — Sat, 26 May 2012 11:55:29 -0700

@Yemon. I didn't highlight that metrics have not to be symmetric because it seems me something implicit in the request. But ok! Now I see that it wasn't clear, and I could edit the OP and stress that "metrics" are not necessarily required to have all the features of standard metrics and norms are not required to have all the features of standard norms, though I don't really know what *you* would identify as the "features of standard norms": Normed groups, valuated domains, vector spaces, C^*-algebras are equally interesting examples of algebraic normed structures, but they have dramatically different characteristics. E.g., suvrit pointed out that vector space norms are convex, but this does not make sense in normed groups. And you observed that *common* normed structured come with a distinguished element whose norm is zero, but this is not the case with normed semigroups. So, at the end of the day, I don't understand how this kind of indirect arguments should prove anything in relation to my question. Apropos, let me repeat it once more: Does there exist any previous attempt to construct a unified framework where metric spaces AND algebraic normed structures can be all regarded as [special] instances of the one same archetypical structure? The term "structure" is formally defined in the OP. Now, what is different from asking: Is there a unified setting to carry out the study, say, of continuity, compactness and connectedness? Would you have considered that as a vague question too? I don't think so. Ok, you would have considered it as a trivial question, but that's not the same. P.S.: In any event, I've just voted up to close the thread.

Yemon Choi comments on "Are the "norms and metrics as parts of some whole" questions appropriate?" (19136)

Yemon Choi — Sat, 26 May 2012 11:10:53 -0700

This question and its predecessor seem too speculative in my view, verging on spurious. But in view of the comment thread filling up I suggest some of the discussion could profitably be moved here.