{"id":4067,"date":"2021-08-24T16:35:39","date_gmt":"2021-08-24T14:35:39","guid":{"rendered":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/?page_id=4067"},"modified":"2022-05-17T09:34:54","modified_gmt":"2022-05-17T07:34:54","slug":"from-prestreams-to-streams","status":"publish","type":"page","link":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/?page_id=4067","title":{"rendered":"From prestreams to streams"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/?page_id=3981\" data-type=\"page\" data-id=\"3981\">Last time<\/a>, I had introduced two models of directed topological spaces, namely of topological spaces with a local direction of time:<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>Marco Grandis&#8217; <em>d-spaces<\/em> [1], which are pairs (<em>X<\/em>, <em>dX<\/em>) of a topological space and of a collection <em>dX<\/em> of paths (namely, continuous maps from [0,1] to <em>X<\/em>), the so-called <em>d-paths<\/em>, that contains all the constant maps, and is closed under concatenation and reparametrization;<\/li><li>Sanjeevi Krishnan&#8217;s <em>prestreams<\/em> [2], which are pairs (<em>X<\/em>, (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>) of a topological space <em>X<\/em> and a <em>precirculation<\/em> (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>, namely a collection of preorderings \u2291<em><sub>U<\/sub><\/em> on <em>U<\/em>, one for each open subset <em>U<\/em> of <em>X<\/em>, with a monotonicity property: if <em>U<\/em> \u2286 <em>V<\/em>, then for all points <em>x<\/em>, <em>y<\/em> of <em>U<\/em> such that <em>x<\/em> \u2291<em><sub>U<\/sub><\/em> <em>y<\/em>, we also have <em>x<\/em> \u2291<em><sub>V<\/sub><\/em> <em>y<\/em>.  (Oops, I had not mentioned the name &#8220;precirculation&#8221; last time; so yes, we are going to call those collections of preorderings precirculations, and there will be things called circulations below.)<\/li><\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">I would like to explain how S. Krishnan&#8217;s notion of stream emerges naturally from the relation between the two notions.  I initially had plans for explaining much more that what I am going to say below, but I am just returning from holidays, and I have not been able to devote the usual 3 to 4 days that it usually takes me to write one of my (now monthly) blog posts.  There is some good in any evil, as one says, and this will allow me to concentrate on just one topic, instead of trying to attack too many questions at once.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Streams<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The real subject of [2] is not prestreams, rather <em>streams<\/em>, which I will talk about today.  Perhaps the simplest way of introducing them is to recall Emmanuel Haucourt&#8217;s <em>S<\/em> \u22a3 <em>D<\/em> adjunction [3], and to find new properties of those prestreams produced by applying the <em>S<\/em> functor to an arbitrary d-space.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The situation is similar with the <strong>O<\/strong> functor from <strong>Top<\/strong> to the opposite of the category <strong>CLat<\/strong> of complete lattices, mapping every topological space to its complete lattice of open sets.  Soon enough we realize that all such lattices obey the so-called frame distributivity law.  We call such lattices <em>frames<\/em>, and we realize that they are interesting in their own right.  Not all frames are obtained by applying <strong>O<\/strong> to a topological space.  Similarly, the interesting property we will find of prestreams obtained by applying <em>S<\/em> to a prestream will turn them into so-called <em>streams<\/em>.  This will be an interesting notion in its own right, although not all streams are obtained by applying <em>S<\/em> to a d-space.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">All right, so how does <em>S<\/em> work?  Given a d-space (<em>X<\/em>, <em>dX<\/em>), we decide to define <em>S<\/em>(<em>X<\/em>, <em>dX<\/em>) as (<em>X<\/em>, (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>), where for every open subset <em>U<\/em> of <em>X<\/em>, as the following preordering on <em>U<\/em>: <em>x<\/em> \u2291<em><sub>U<\/sub><\/em> <em>y<\/em> if and only if there is a d-path \u03b3 from <em>x<\/em> to <em>y<\/em> (i.e., \u03b3(0)=<em>x<\/em> and \u03b3(1)=<em>y<\/em>), whose image Im \u03b3 is entirely included in <em>U<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Fact.<\/strong>  Let (<em>X<\/em>, (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>) \u225d <em>S<\/em>(<em>X<\/em>, <em>dX<\/em>).  Then (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub> is a <em>circulation<\/em>, namely: for every open subset <em>U<\/em> of <em>X<\/em>, for every open cover <strong>C<\/strong> of the subspace <em>U<\/em> (namely, each <em>V<\/em> \u2208 <strong>C<\/strong> is open in <em>X<\/em> and included in <em>U<\/em>), \u2291<em><sub>U<\/sub><\/em> coincides with the &#8220;join&#8221; of the preorderings \u2291<em><sub>V<\/sub><\/em>, <em>V<\/em> \u2208 <strong>C<\/strong>, in the sense that:<\/p>\n\n\n\n<p class=\"has-text-align-center wp-block-paragraph\">for all points <em>x<\/em>, <em>y<\/em> in <em>U<\/em>, <em>x<\/em> \u2291<em><sub>U<\/sub><\/em> <em>y<\/em> if and only if there are finitely many elements <em>V<\/em><sub>1<\/sub>, &#8230;, <em>V<sub>n<\/sub><\/em> of <strong>C<\/strong> and matching elements <em>x<\/em><sub>0<\/sub>, <em>x<\/em><sub>1<\/sub>, &#8230;, <em>x<sub>n<\/sub><\/em> such that <em>x<\/em>=<em>x<\/em><sub>0<\/sub> \u2291<em><sub>V<\/sub><\/em><sub>1<\/sub> <em>x<\/em><sub>1<\/sub> \u2291<em><sub>V<\/sub><\/em><sub>2<\/sub> &#8230; \u2291<em><sub>V<\/sub><\/em><sub><em>n<\/em><\/sub> <em>x<sub>n<\/sub><\/em>=<em>y<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The latter formulation implicitly entails that <em>x<\/em><sub>0<\/sub> and <em>x<\/em><sub>1<\/sub> are both in <em>V<\/em><sub>1<\/sub>, <em>x<\/em><sub>1<\/sub> and <em>x<\/em><sub>2<\/sub> are both in <em>V<\/em><sub>2<\/sub>, &#8230;, <em>x<\/em><sub><em>n<\/em>\u20131<\/sub> and <em>x<\/em><sub><em>n<\/em><\/sub> are both in <em>V<\/em><sub><em>n<\/em><\/sub>.  I will explain what this &#8220;join&#8221; is, really, later.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>Proof.<\/em>  This is a consequence of the compactness of [0,1], really, plus a certain form of connectedness.  Let me recall that we equip [0,1] with its usual Hausdorff topology; not with its Scott topology, for example.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The if direction is trivial: if <em>x<\/em>=<em>x<\/em><sub>0<\/sub> \u2291<em><sub>V<\/sub><\/em><sub>1<\/sub> <em>x<\/em><sub>1<\/sub> \u2291<em><sub>V<\/sub><\/em><sub>2<\/sub> &#8230; \u2291<em><sub>V<\/sub><\/em><sub><em>n<\/em><\/sub> <em>x<sub>n<\/sub><\/em>=<em>y<\/em>, we must have <em>x<\/em>=<em>x<\/em><sub>0<\/sub> \u2291<em><sub>U<\/sub><\/em> <em>x<\/em><sub>1<\/sub> \u2291<em><sub>U<\/sub><\/em> &#8230; \u2291<em><sub>U<\/sub><\/em> <em>x<sub>n<\/sub><\/em>=<em>y<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Conversely, we use the fact that (<em>X<\/em>, (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>) is equal to <em>S<\/em>(<em>X<\/em>, <em>dX<\/em>): hence there is a d-path \u03b3 : [0,1] \u2192&nbsp;<em>X<\/em> such that \u03b3(0)=<em>x<\/em>, \u03b3(1)=<em>y<\/em>, and Im \u03b3 \u2286 <em>U<\/em>.  The situation is as follows:<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large is-resized\"><a href=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams1.png\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams1.png\" alt=\"\" class=\"wp-image-4083\" width=\"341\" height=\"137\"\/><\/a><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">Also, <em>U<\/em> = \u222a<strong>C<\/strong>, so <em>U<\/em> really is a union of open subsets, which we display as the following collection of darker blue potato shapes; each one is an open set <em>V<\/em> from the cover <strong>C<\/strong>.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large is-resized\"><a href=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams2-1.png\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams2-1.png\" alt=\"\" class=\"wp-image-4090\" width=\"341\" height=\"137\"\/><\/a><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">Then \u03b3<sup>\u20131<\/sup>(<em>U<\/em>) = \u222a<sub><em>V<\/em> \u2208 <strong>C<\/strong><\/sub> \u03b3<sup>\u20131<\/sup>(<em>V<\/em>), and therefore the sets \u03b3<sup>\u20131<\/sup>(<em>V<\/em>), <em>V<\/em> \u2208 <strong>C<\/strong>, form an open cover of [0,1].<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The desired sets <em>V<\/em><sub>1<\/sub>, &#8230;, <em>V<sub>n<\/sub><\/em> are obtained by extracting a finite subcover of that open cover of [0,1], as you may guess.  In pictures, here is what we are looking for:<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large is-resized\"><a href=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams3-1.png\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams3-1.png\" alt=\"\" class=\"wp-image-4097\" width=\"342\" height=\"147\"\/><\/a><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">Let us get on with the technical details.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">A base of the topology of [0,1] is given by what I will call <em>open intervals<\/em>: those are the intersections of actual open intervals of <strong>R<\/strong> with [0,1], hence are sets of the form ]<em>a<\/em>,<em>b<\/em>[ with 0\u2264<em>a<\/em>&lt;<em>b<\/em>\u22641, and also sets of the form [0,<em>b<\/em>[ with 0&lt;<em>b<\/em>\u22641, and ]<em>a<\/em>,1] with 0\u2264<em>a<\/em>&lt;1.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">For each <em>t<\/em> \u2208 [0,1], <em>t<\/em> is in \u03b3<sup>\u20131<\/sup>(<em>V<\/em>) for some <em>V<\/em> \u2208 <strong>C<\/strong>, and since \u03b3<sup>\u20131<\/sup>(<em>V<\/em>) is open, <em>t<\/em> must be contained in some open interval <em>I<\/em>[<em>t<\/em>] contained in that set \u03b3<sup>\u20131<\/sup>(<em>V<\/em>).  The open sets <em>I<\/em>[<em>t<\/em>], <em>t<\/em> \u2208 [0,1], themselves form an open cover of [0,1], and we extract a finite subcover <em>I<\/em>[<em>t<\/em><sub>1<\/sub>], &#8230;, <em>I<\/em>[<em>t<\/em><sub><em>m<\/em><\/sub>].  Here is how this finite subcover, consisting of open intervals, may look like.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large is-resized\"><a href=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams-I1.png\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams-I1.png\" alt=\"\" class=\"wp-image-4100\" width=\"289\" height=\"121\"\/><\/a><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\"> We massage the latter list in order to ensure that none of the latter intervals contains any other, by removing any <em>I<\/em>[<em>t<\/em><sub><em>j<\/em><\/sub>] that is included in some <em>I<\/em>[<em>t<\/em><sub><em>k<\/em><\/sub>], for some <em>k<\/em>\u2260<em>j<\/em>, iteratively.  For example, there are two intervals containing 0 in the picture above, and we keep only the larger of the two.  In pictures, we obtain the following.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large is-resized\"><a href=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams-I2.png\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams-I2.png\" alt=\"\" class=\"wp-image-4102\" width=\"288\" height=\"121\"\/><\/a><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">The result is a list of pairwise incomparable intervals, with respect to inclusion.  We can then sort this list from left to right.  This means that the left endpoint of <em>I<\/em>[<em>t<\/em><sub>1<\/sub>] is smaller than or equal to the left endpoint of <em>I<\/em>[<em>t<\/em><sub>2<\/sub>], which is smaller than or equal to the left endpoint of <em>I<\/em>[<em>t<\/em><sub>3<\/sub>], etc., and similarly for right endpoints.  This is how we sorted them in the picture above.  Formally, this rests on the following observation: given any two intervals with endpoints <em>a<\/em>,<em>b<\/em> and <em>c<\/em>,<em>d<\/em> respectively that are incomparable with respect to inclusion, either <em>a<\/em>\u2264<em>c<\/em> and <em>b<\/em>\u2264<em>d<\/em>, or <em>a<\/em>\u2265<em>c<\/em> and <em>b<\/em>\u2265<em>d<\/em>, namely comparing them by left endpoints <em>a<\/em>,<em>c<\/em> gives the same result as comparing them by right endpoints <em>b<\/em>,<em>d<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">For each <em>j<\/em> with 1\u2264<em>j<\/em>\u2264<em>m<\/em>, let <em>a<sub>j<\/sub><\/em> be the left endpoint of <em>I<\/em>[<em>t<\/em><sub><em>j<\/em><\/sub>] and let <em>b<sub>j<\/sub><\/em> be its right endpoint.  Since 0\u2264<em>a<\/em><sub>1<\/sub>\u2264<em>a<\/em><sub>2<\/sub>\u2264&#8230;\u2264<em>a<\/em><sub><em>m<\/em><\/sub>\u22641, and since <em>I<\/em>[<em>t<\/em><sub>1<\/sub>], &#8230;, <em>I<\/em>[<em>t<\/em><sub><em>m<\/em><\/sub>] cover [0,1], we must have <em>a<\/em><sub>1<\/sub>=0.  Similarly, <em>b<\/em><sub><em>m<\/em><\/sub>=1.  Also, successive intervals <em>I<\/em>[<em>t<\/em><sub><em>j<\/em><\/sub>] and <em>I<\/em>[<em>t<\/em><sub><em>j<\/em>+1<\/sub>] must overlap: otherwise no point in the open interval ]<em>b<\/em><sub><em>j<\/em><\/sub>,<em>a<\/em><sub><em>j+1<\/em><\/sub>[ would belong to any <em>I<\/em>[<em>t<\/em><sub><em>k<\/em><\/sub>]: those with <em>k<\/em>\u2264<em>j<\/em> cover only points that lie to the left of <em>b<\/em><sub><em>j<\/em><\/sub>, and those with <em>k<\/em>\u2265<em>j<\/em>+1 cover only points that lie to the right of <em>a<\/em><sub><em>j<\/em>+1<\/sub>.  Since they overlap, we pick an element <em>s<\/em><sub><em>j<\/em><\/sub> in <em>I<\/em>[<em>t<\/em><sub><em>j<\/em><\/sub>] and <em>I<\/em>[<em>t<\/em><sub><em>j<\/em>+1<\/sub>], for each <em>j<\/em> such that 1\u2264<em>j<\/em>\u2264<em>m<\/em>\u20131.  Let us also define <em>s<\/em><sub>0<\/sub> be 0, and <em>s<\/em><sub><em>m<\/em><\/sub> be 1.  The situation is now as follows.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large is-resized\"><a href=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams-I3.png\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams-I3.png\" alt=\"\" class=\"wp-image-4104\" width=\"289\" height=\"121\"\/><\/a><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">Let us look at the list of elements 0=<em>s<\/em><sub>0<\/sub>\u2264<em>s<\/em><sub>1<\/sub>\u2264<em>s<\/em><sub>2<\/sub>\u2264&#8230;\u2264<em>s<\/em><sub><em>m<\/em><\/sub>=1.  It is easy to check that [<em>s<\/em><sub>0<\/sub>,<em>s<\/em><sub>1<\/sub>] is included in <em>I<\/em>[<em>t<\/em><sub>1<\/sub>], [<em>s<\/em><sub>1<\/sub>,<em>s<\/em><sub>2<\/sub>] is included in <em>I<\/em>[<em>t<\/em><sub>2<\/sub>], &#8230;, [<em>s<\/em><sub><em>m<\/em>\u20131<\/sub>,<em>s<\/em><sub><em>m<\/em><\/sub>] is included in <em>I<\/em>[<em>t<\/em><sub><em>m<\/em><\/sub>].  We define <em>n<\/em> as <em>m<\/em>+1, and the open subsets <em>V<\/em><sub>1<\/sub>, &#8230;, <em>V<sub>n<\/sub><\/em> of <em>U<\/em> as follows.  For each <em>j<\/em>, 1\u2264<em>j<\/em>\u2264<em>n<\/em>=<em>m<\/em>+1, [<em>s<\/em><sub><em>j<\/em>\u20131<\/sub>,<em>s<\/em><sub><em>j<\/em><\/sub>] is included in <em>I<\/em>[<em>t<\/em><sub><em>j<\/em><\/sub>], and the latter is included in an open set \u03b3<sup>\u20131<\/sup>(<em>V<\/em>) for some <em>V<\/em> \u2208 <strong>C<\/strong>: we let <em>V<\/em><sub><em>j<\/em><\/sub> be any such <em>V<\/em>.  The situation is now as follows.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large is-resized\"><a href=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams4.png\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2021\/08\/streams4.png\" alt=\"\" class=\"wp-image-4098\" width=\"343\" height=\"147\"\/><\/a><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">We now let <em>x<sub>j<\/sub><\/em> \u225d \u03b3(<em>s<\/em><sub><em>j<\/em><\/sub>) for each <em>j<\/em>, 0\u2264<em>j<\/em>\u2264<em>n<\/em>.  For each <em>j<\/em> with 1\u2264<em>j<\/em>\u2264<em>n<\/em>, \u03b3 has an obvious subpath from <em>x<\/em><sub><em>j<\/em>\u20131<\/sub> to <em>x<\/em><sub><em>j<\/em><\/sub>, which one obtains by reparametrization.  In particular, that subpath is a d-path.  Since [<em>s<\/em><sub><em>j<\/em>\u20131<\/sub>,<em>s<\/em><sub><em>j<\/em><\/sub>] \u2286 <em>I<\/em>[<em>t<\/em><sub><em>j<\/em><\/sub>] \u2286 \u03b3<sup>\u20131<\/sup>(<em>V<\/em><sub><em>j<\/em><\/sub>), the image of that subpath is entirely included in <em>V<\/em><sub><em>j<\/em><\/sub>.  Therefore <em>x<\/em><sub><em>j<\/em>\u20131<\/sub> \u2291<em><sub>V<\/sub><\/em><sub><em>j<\/em><\/sub> <em>x<\/em><sub><em>j<\/em><\/sub>.  \u2610<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">A prestream (<em>X<\/em>, (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>) whose precirculation (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub> is a circulation (namely satisfies the condition of the above Fact) is called a <em>stream<\/em>.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">A few examples, a few non-examples<\/h2>\n\n\n\n<ul class=\"wp-block-list\" id=\"block-648dee79-52bf-40f4-af69-77b594794d1b\"><li>Recall the directed circle, defined as the following prestream <em>S<\/em><sub>1<\/sub><sup>\u2192<\/sup>: this is the space of complex numbers of modulus 1, namely the numbers of the form e<sup><em>i<\/em>\u03b8<\/sup>, \u03b8 \u2208 <strong>R<\/strong>, with the subspace topology induced by the inclusion in <strong>C<\/strong>, as in the d-space case.<br>For every open subset <em>U<\/em>, for all <em>x<\/em>, <em>y<\/em> in <em>U<\/em>, <em>x<\/em> \u2291<em><sub>U<\/sub><\/em> <em>y<\/em> holds if and only if we can write <em>x<\/em> as e<sup><em>i<\/em>\u03b8<\/sup>, <em>y<\/em> as e<sup><em>i<\/em>\u03b8&#8217;<\/sup>, with \u03b8\u2264\u03b8&#8217;, <em>and<\/em> such that for every <em>t<\/em> in the interval [\u03b8, \u03b8&#8217;], e<sup><em>it<\/em><\/sup> is in <em>U<\/em>.  We had seen last time that this is a Haucourt stream, namely the image by the functor <em>S<\/em> of a certain d-space.  As such, <em>S<\/em><sub>1<\/sub><sup>\u2192<\/sup> is a stream.  In fact, every Haucourt stream is a stream.<\/li><li>The&nbsp;directed line segment&nbsp;<strong>I<\/strong><sup>\u2192<\/sup> is a Haucourt stream, hence a stream. Let me recall that its (pre)circulation is defined as follows: for every open subset&nbsp;<em>U<\/em>&nbsp;of [0,1], for all&nbsp;<em>x<\/em>,&nbsp;<em>y<\/em>&nbsp;in&nbsp;<em>U<\/em>,&nbsp;<em>x<\/em>&nbsp;\u2291<em><sub>U<\/sub><\/em>&nbsp;<em>y<\/em>&nbsp;if and only if&nbsp;<em>x<\/em>\u2264<em>y<\/em> and&nbsp;the&nbsp;<em>whole segment<\/em>&nbsp;[<em>x<\/em>,<em>y<\/em>] is included in&nbsp;<em>U<\/em>.<\/li><li>Every topological space <em>X<\/em> equipped with a preordering \u2264 defines a prestream by declaring that \u2291<em><sub>U<\/sub><\/em> is the restriction of \u2264 to <em>U<\/em> for every open subset <em>U<\/em> of <em>X<\/em>.  This is almost never a stream.  For that, you would need the circulation property to hold, namely &#8220;for every open subset <em>U<\/em> of <em>X<\/em>, for every open cover <strong>C<\/strong> of the subspace <em>U<\/em>, for all points <em>x<\/em>, <em>y<\/em> in <em>U<\/em>, if <em>x<\/em> \u2264 <em>y<\/em> then there are finitely many elements <em>V<\/em><sub>1<\/sub>, &#8230;, <em>V<sub>n<\/sub><\/em> of <strong>C<\/strong> and matching elements <em>x<\/em><sub>0<\/sub>, <em>x<\/em><sub>1<\/sub>, &#8230;, <em>x<sub>n<\/sub><\/em> such that <em>x<\/em>=<em>x<\/em><sub>0<\/sub> \u2264 <em>x<\/em><sub>1<\/sub> \u2264 &#8230; \u2264 <em>x<sub>n<\/sub><\/em>=<em>y<\/em>, and each pair <em>x<\/em><sub><em>i<\/em>\u20131<\/sub>, <em>x<\/em><sub><em>i<\/em><\/sub> lies in <em>V<sub>i<\/sub><\/em>&#8220;.  (Do not forget the last part of the condition: when we say that <em>x<\/em><sub><em>i<\/em>\u20131<\/sub> is below <em>x<\/em><sub><em>i<\/em><\/sub> in the restriction of \u2264 to <em>V<sub>i<\/sub><\/em>, this really means that not only <em>x<\/em><sub><em>i<\/em>\u20131<\/sub> \u2264 <em>x<\/em><sub><em>i<\/em><\/sub>, but also that both <em>x<\/em><sub><em>i<\/em>\u20131<\/sub> and <em>x<\/em><sub><em>i<\/em><\/sub> are in <em>V<sub>i<\/sub><\/em>.)<br>For example, this condition fails if you take [0,1] for <em>X<\/em>, and its usual ordering for \u2264: let <em>U<\/em> be the disjoint union of [0, 1\/3[ and ]2\/3, 1], <em>x<\/em> be 0, and <em>y<\/em> be 1, and let the cover consist of just the two open sets [0, 1\/3[ and ]2\/3, 1].  Imagine that you could find <em>V<\/em><sub>1<\/sub>, &#8230;, <em>V<sub>n<\/sub><\/em> in the cover and elements <em>x<\/em><sub>0<\/sub>, <em>x<\/em><sub>1<\/sub>, &#8230;, <em>x<sub>n<\/sub><\/em> as above.  (By the way, we only have two sets in that cover, but the list <em>V<\/em><sub>1<\/sub>, &#8230;, <em>V<sub>n<\/sub><\/em> can, and will usually, contain repeated copies of those two sets; I never insisted that <em>V<\/em><sub>1<\/sub>, &#8230;, <em>V<sub>n<\/sub><\/em> be <em>distinct<\/em> open sets, right?)  Now <em>x<\/em>=<em>x<\/em><sub>0<\/sub> is in [0, 1\/3[ and not in ]2\/3, 1], so <em>V<\/em><sub>1<\/sub> must be [0, 1\/3[.  Next, the pair <em>x<\/em><sub>0<\/sub>, <em>x<\/em><sub>1<\/sub> lies in <em>V<\/em><sub>1<\/sub>, so <em>x<\/em><sub>1<\/sub> is also in [0, 1\/3[.  Then the pair <em>x<\/em><sub>1<\/sub>, <em>x<\/em><sub>2<\/sub> lies in <em>V<\/em><sub>1<\/sub>, so <em>x<\/em><sub>2<\/sub> is also in [0, 1\/3[.  Continuing in the same way, we realize that every <em>x<sub>i<\/sub><\/em> must be in [0, 1\/3[.  This is in particular true for <em>x<sub>n<\/sub><\/em>=<em>y<\/em>=1, which is absurd.<\/li><li>Every topological space can be given the&nbsp;<em>discrete<\/em>&nbsp;precirculation, where&nbsp;\u2291<em><sub>U<\/sub><\/em>&nbsp;is the trivial preordering on&nbsp;<em>U<\/em>&nbsp;that makes every point below any other point. &nbsp;This is a special case of a preordered topological space (see previous bulleted item), where the preordering is the trivial one.<br>By a similar argument as above, the discrete precirculation cannot be a circulation if <em>X<\/em> contains a disconnected open set <em>U<\/em>, in other words if <em>U<\/em> can be written as a disjoint union of two non-empty open subsets.  This rules out a lot of topological spaces!  However, if <em>X<\/em> has a top element \u22a4 in its specialization preordering, then the discrete precirculation <em>is<\/em> a circulation: given any two points <em>x<\/em> and <em>y<\/em> of an open set <em>U<\/em>, and an open cover <strong>C<\/strong> of the subspace <em>U<\/em>, we pick one element <em>V<\/em><sub>1<\/sub> from <strong>C<\/strong> that contains <em>x<\/em>, one element <em>V<\/em><sub>2<\/sub> from <strong>C<\/strong> that contains <em>y<\/em>, and we realize that <em>x<\/em> \u2291<em><sub>V<\/sub><\/em><sub>1<\/sub> \u22a4 \u2291<em><sub>V<\/sub><\/em><sub>2<\/sub>  <em>y<\/em> (recall that \u2291<em><sub>V<\/sub><\/em><sub>1<\/sub> and \u2291<em><sub>V<\/sub><\/em><sub>2<\/sub> are both trivial), that the pair <em>x<\/em>, \u22a4 lies in <em>V<\/em><sub>1<\/sub>, and that the pair \u22a4, <em>y<\/em> lies in <em>V<\/em><sub>2<\/sub>.<\/li><li>Every topological space can be given the&nbsp;<em>indiscrete<\/em>&nbsp;precirculation, where&nbsp;\u2291<em><sub>U<\/sub><\/em>&nbsp;is just equality. &nbsp;This is again a preordered topological space, with equality as preordering.  This one is <em>always<\/em> a circulation: if <em>x<\/em> \u2291<em><sub>U<\/sub><\/em> <em>y<\/em>, then in fact <em>x<\/em>=<em>y<\/em>, and from there the circulation property is trivial.<\/li><\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">A lattice-theoretic view<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Earlier, I mentioned that a precirculation (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub> is a circulation if \u2291<em><sub>U<\/sub><\/em> coincides with the &#8220;join&#8221; of the preorderings \u2291<em><sub>V<\/sub><\/em>, <em>V<\/em> \u2208 <strong>C<\/strong>, for every open subset <em>U<\/em> of <em>X<\/em> and every open cover <strong>C<\/strong> of the subspace <em>U<\/em>.  I also promised that I would explain the term &#8220;join&#8221;.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The typical view is that this condition is equivalent to requiring that the preordered set (<em>U<\/em>, \u2291<em><sub>U<\/sub><\/em>) be the colimit (up to isomorphism), taken in the category of preordered sets and monotonic maps, of the diagram of preordered sets (<em>V<\/em>, \u2291<em><sub>V<\/sub><\/em>), where <em>V<\/em> ranges over the open subsets of <em>U<\/em>, and with the obvious inclusions as morphisms.  (I am saying that out of memory, and I have not checked it again, so please excuse me if that is incorrect.)  Hence &#8220;join&#8221; means colimit.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">This is a nice observation, and one that leads to the view of streams as cosheaves, which is most certainly how S. Krishnan came to invent them [2].<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">However, this view misses an important point, and one that will stare you in the face if you ever try to design a Stone duality theory for streams (something I planned to do, but that I have not completed yet): the <em>points<\/em> of the preordered set at index <em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em> should be the points of <em>U<\/em> itself (up to isomorphism).  We are not defining precirculations as any old diagram of preordered sets indexed by the open subsets <em>U<\/em> of <em>X<\/em>; the set of points in those preordered sets must be the set of points of the index <em>U<\/em> itself.  This is a pretty strong requirement, really.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">I cannot resist mentioning another formulation of prestreams and streams, which I have not seen in the literature, and which avoids this pitfall entirely.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Let call a <em>modified prestream<\/em> any pair (<em>X<\/em>, (\u2291&#8217;<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>) of a topological space <em>X<\/em>, and of a family of preorderings \u2291&#8217;<em><sub>U<\/sub><\/em> on <em>X<\/em> (not on <em>U<\/em>! this is the main modification) for each open subset <em>U<\/em> of <em>X<\/em> such that:<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>the usual monotonicity condition holds: if <em>x<\/em> \u2291&#8217;<em><sub>U<\/sub><\/em> <em>y<\/em> and <em>U<\/em> is included in <em>V<\/em>, then <em>x<\/em> \u2291&#8217;<em><sub>V<\/sub><\/em> <em>y<\/em>;<\/li><li>for each open subset <em>U<\/em> of <em>X<\/em>, \u2291&#8217;<em><sub>U<\/sub><\/em> is <em>supported<\/em> on <em>U<\/em>, namely: for all points <em>x<\/em> \u2291&#8217;<em><sub>U<\/sub><\/em> <em>y<\/em> in <em>X<\/em>, <em>x<\/em> and <em>y<\/em> are both in <em>U<\/em> or <em>x<\/em>=<em>y<\/em>.  This may be clearer if I state that in the equivalent way: for all <em>distinct<\/em> points <em>x<\/em> \u2291&#8217;<em><sub>U<\/sub><\/em> <em>y<\/em>, <em>x<\/em> and <em>y<\/em> must be in <em>U<\/em>; or, equivalently again: for all points <em>x<\/em> \u2291&#8217;<em><sub>U<\/sub><\/em> <em>y<\/em>, if <em>x<\/em> or <em>y<\/em> is not in <em>U<\/em> then <em>x<\/em>=<em>y<\/em>.<\/li><\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">Given any prestream (<em>X<\/em>, (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>), we can define an associated modified prestream (<em>X<\/em>, (\u2291&#8217;<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>) by letting <em>x<\/em> \u2291&#8217;<em><sub>U<\/sub><\/em> <em>y<\/em> hold if and only if either both <em>x<\/em> and <em>y<\/em> are in <em>U<\/em> and <em>x<\/em> \u2291<em><sub>U<\/sub><\/em> <em>y<\/em>, or <em>x<\/em>=<em>y<\/em> is not in <em>U<\/em>.  (\u2291&#8217;<em><sub>U<\/sub><\/em> is the <em>modification<\/em> of \u2291<em><sub>U<\/sub><\/em>.)  Conversely, given any modified prestream (<em>X<\/em>, (\u2291&#8217;<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>), we obtain a prestream (<em>X<\/em>, (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>) by declaring that \u2291<em><sub>U<\/sub><\/em> is the restricting of \u2291&#8217;<em><sub>U<\/sub><\/em> to <em>U<\/em>, for each open subset <em>U<\/em> of <em>X<\/em>.  The two constructions are inverse of each other.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">There is a category <strong>PreStr&#8217;<\/strong> of modified prestreams, whose morphisms <em>f<\/em>&nbsp;: (<em>X<\/em>, (\u2291&#8217;<em><sub>U<\/sub><\/em>)<sub><em>U<\/em>&nbsp;\u2208&nbsp;<strong>O<\/strong><em>X<\/em><\/sub>) \u2192 (<em>Y<\/em>, (\u2aaf&#8217;<em><sub>V<\/sub><\/em>)<sub><em>V<\/em>&nbsp;\u2208&nbsp;<strong>O<\/strong><em>Y<\/em><\/sub>) are the continuous maps <em>f<\/em> from <em>X<\/em> to <em>Y<\/em> that are monotonic from the preordered set (<em>X<\/em>, \u2291&#8217;<em><sub>U<\/sub><\/em>) to (<em>Y<\/em>, \u2aaf&#8217;<em><sub>V<\/sub><\/em>), where <em>U<\/em> and <em>V<\/em> range over the pairs of open subsets such that <em>U<\/em>=<em>f<\/em><sup>\u20131<\/sup>(<em>V<\/em>).  (Yes, that means an infinite number of monotonicity conditions.)<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">It is easy to see that <strong>PreStr<\/strong> and <strong>PreStr&#8217;<\/strong> are equivalent categories\u2014in fact <em>isomorphic<\/em> categories; we have just described the natural isomorphism between the two categories.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Now there is yet another formulation of modified prestreams.  There is a complete lattice of preorderings <strong>Pre<\/strong>(<em>X<\/em>) on the set <em>X<\/em>.  The lattice structure is as follows.  Preorderings are compared by inclusion: \u2291&#8217; is included in \u2aaf&#8217; if and only if for all points <em>x<\/em> and <em>y<\/em> such that <em>x<\/em> \u2291&#8217; <em>y<\/em>, we also have \u2aaf&#8217;.  Then infima are just intersections: <em>x<\/em> is below <em>y<\/em> with respect to the infimum of a family of preorderings if and only if <em>x<\/em> is below <em>y<\/em> with respect to each of the preorderings in the family.  And since we have all infima, we have all suprema, too: the supremum of a family is the infimum of its set of upper bounds.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">One can check that, explicitly, the supremum \u2291&#8217; of a family of preorderings (\u2291&#8217;<em><sub>i<\/sub><\/em>)<sub><em>i<\/em>&nbsp;\u2208&nbsp;<em>I<\/em><\/sub> on <em>X<\/em> is given by <em>x<\/em> \u2291&#8217; <em>y<\/em> if and only if there are finitely many indices <em>i<\/em><sub>1<\/sub>, &#8230;, <em>i<sub>n<\/sub><\/em> in <em>I<\/em>, and a matching sequence of points <em>x<\/em>=<em>x<\/em><sub>0<\/sub> \u2291<em><sub>i<\/sub><\/em><sub>1<\/sub> <em>x<\/em><sub>1<\/sub> \u2291<em><sub>i<\/sub><\/em><sub>2<\/sub> &#8230; \u2291<em><sub>i<\/sub><\/em><sub><em>n<\/em><\/sub> <em>x<sub>n<\/sub><\/em>=<em>y<\/em>.  (I guess you can see the connection with circulations emerge now.)<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">For every (open, although that is not essential) subset <em>U<\/em> of <em>X<\/em>, there is a largest preordering \u2291&#8217;<sup>0<\/sup><em><sub>U<\/sub><\/em> on <em>X<\/em> that is supported on <em>U<\/em>.  This is the preordering that declares <em>x<\/em> to be below <em>y<\/em> if and only if either both <em>x<\/em> and <em>y<\/em> are in <em>U<\/em> (without any further condition on <em>U<\/em>), or <em>x<\/em>=<em>y<\/em>.  This defines a map <strong>d<\/strong><em><sub>X<\/sub><\/em> from <strong>O<\/strong><em>X<\/em> to <strong>Pre<\/strong>(<em>X<\/em>), which sends <em>U<\/em> to \u2291&#8217;<sup>0<\/sup><em><sub>U<\/sub><\/em>.  (&#8220;<strong>d<\/strong>&#8221; is for &#8220;discrete&#8221;.)  This map is monotonic, and a preordering \u2291&#8217; on <em>X<\/em> is supported on <em>U<\/em> if and only if \u2291&#8217; is included in <strong><strong>d<\/strong><em><sub>X<\/sub><\/em><\/strong>(<em>U<\/em>)=\u2291&#8217;<sup>0<\/sup><em><sub>U<\/sub><\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Hence a modified prestream is, equivalently, a pair of a topological space <em>X<\/em> and of a monotonic map <em>U<\/em> \u21a6 \u2291&#8217;<em><sub>U<\/sub><\/em> from <strong>O<\/strong><em>X<\/em> to <strong>Pre<\/strong>(<em>X<\/em>) that is (pointwise) below the map <strong><strong>d<\/strong><em><sub>X<\/sub><\/em><\/strong>.  In other words, it is a pair (<em>X<\/em>, <em>circ<\/em>) where <em>circ<\/em> is an element of \u2193<strong>d<\/strong><em><sub>X<\/sub><\/em>, the downward closure of <strong>d<\/strong><em><sub>X<\/sub><\/em> in the complete lattice of monotonic maps from <strong>O<\/strong><em>X<\/em> to <strong>Pre<\/strong>(<em>X<\/em>).  Through the isomorphism between <strong>PreStr<\/strong> and <strong>PreStr&#8217;<\/strong>, such elements of \u2193<strong>d<\/strong><em><sub>X<\/sub><\/em> are exactly the precirculations on <em>X<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Now what is a circulation?  That it is simply an element of \u2193<strong>d<\/strong><em><sub>X<\/sub><\/em> that <em>preserves all suprema<\/em> (up to the isomorphism between unmodified and modified precirculations).<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Indeed, one can characterize circulations without relying on covers in the following way: a precirculation (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em>&nbsp;\u2208&nbsp;<strong>O<\/strong><em>X<\/em><\/sub> on <em>X<\/em> is a circulation if and only if for every family <strong>C<\/strong> of open subsets of <em>X<\/em>, letting <em>U<\/em>\u225dsup <strong>C<\/strong> (the union of all the elements of <strong>C<\/strong>), \u2291<em><sub>U<\/sub><\/em> is the preordering on <em>X<\/em> defined by the fact that for all points <em>x<\/em>, <em>y<\/em> in <em>U<\/em>, <em>x<\/em> \u2291<em><sub>U<\/sub><\/em> <em>y<\/em> if and only if there are finitely many elements <em>V<\/em><sub>1<\/sub>, &#8230;, <em>V<sub>n<\/sub><\/em> of <strong>C<\/strong> and matching elements <em>x<\/em><sub>0<\/sub>, <em>x<\/em><sub>1<\/sub>, &#8230;, <em>x<sub>n<\/sub><\/em> such that <em>x<\/em>=<em>x<\/em><sub>0<\/sub> \u2291<em><sub>V<\/sub><\/em><sub>1<\/sub> <em>x<\/em><sub>1<\/sub> \u2291<em><sub>V<\/sub><\/em><sub>2<\/sub> &#8230; \u2291<em><sub>V<\/sub><\/em><sub><em>n<\/em><\/sub> <em>x<sub>n<\/sub><\/em>=<em>y<\/em>.  Relying on the explicit characterization of suprema in <strong>Pre<\/strong>(<em>X<\/em>) that I gave above, you will see that this means that the modification \u2291&#8217;<em><sub>U<\/sub><\/em> of \u2291<em><sub>U<\/sub><\/em> is required to be the supremum of the modifications \u2291&#8217;<em><sub>V<\/sub><\/em> of the preorderings \u2291<em><sub>V<\/sub><\/em>, <em>V<\/em> \u2208 <strong>C<\/strong>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">In other words:<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\"><p>A (modified) stream is a pair (<em>X<\/em>, <em>circ<\/em>) where <em>circ<\/em> is an element of the complete lattice of <em>supremum preserving<\/em> maps from <strong>O<\/strong><em>X<\/em> to <strong>Pre<\/strong>(<em>X<\/em>), below <strong>d<\/strong><em><sub>X<\/sub><\/em>.<\/p><\/blockquote>\n\n\n\n<p class=\"wp-block-paragraph\">(In passing, beware that <strong>d<\/strong><em><sub>X<\/sub><\/em> itself is not in general supremum preserving.  The argument is exactly the same as the one that we used to show that the discrete precirculation on [0,1] is not a circulation.)<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">I was somehow hoping that this observation would perhaps lead to a nice theory of Stone duality for streams.  One would consider general frames <em>L<\/em> (instead of <strong>O<\/strong><em>X<\/em>) together with a supremum preserving map from <em>L<\/em> to some localic stand-in for \u2193<strong>d<\/strong><em><sub>X<\/sub><\/em> \u2286 <strong>Pre<\/strong>(<em>X<\/em>).  The latter is the problem: in general, <strong>Pre<\/strong>(<em>X<\/em>) is very far from being a frame!  In fact, it is in general not even distributive, as I will let you check for yourselves.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Well, another time, maybe.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">The category of streams<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">There is a full subcategory <strong>Str<\/strong> of prestreams that consists of streams.  Its morphisms are exactly the stream morphisms, namely the continuous, locally monotonic maps.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The adjunction <em>S<\/em> \u22a3 <em>D<\/em> restricts to one between the category <strong>dTop<\/strong> of d-spaces and <strong>Str<\/strong>: by definition, <em>S<\/em> maps into <strong>Str<\/strong>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">One can show that <strong>Str<\/strong> is complete, cocomplete, and that colimits are computed exactly as in the larger category <strong>PreStr<\/strong> of prestreams.  Hence in particular, the directed circle <em>S<\/em><sub>1<\/sub><sup>\u2192<\/sup> is also a quotient of the directed line segment <strong>I<\/strong><sup>\u2192<\/sup> in <strong>Str<\/strong>, where by quotient I mean the relevant coequalizer, if that does not frighten you.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Limits are very different.  One can show that <strong>Str<\/strong> is a coflective subcategory of <strong>PreStr<\/strong>, namely that the inclusion functor (of <strong>PreStr<\/strong> into <strong>Str<\/strong>) has a right adjoint (see [2], or Corollary 3 in [4]).  This right adjoint is called <em>cosheafification<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Intuitively, the cosheafification of a prestream (<em>X<\/em>, (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>) is obtained as follows.<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>We first replace each preordering \u2291<em><sub>U<\/sub><\/em> by a new one ^\u2291<em><sub>U<\/sub><\/em> (the hat ^ is written above \u2291 in [4], but this is hard to render in html), defined so as to repair the circulation condition.  Namely, we define <em>x<\/em> ^\u2291<em><sub>U<\/sub><\/em> <em>y<\/em>, for all points <em>x<\/em> and <em>y<\/em> of an arbitrary open set <em>U<\/em> as: for every open cover <strong>C<\/strong> of the subspace <em>U<\/em>, there are finitely many elements <em>V<\/em><sub>1<\/sub>, &#8230;, <em>V<sub>n<\/sub><\/em> of <strong>C<\/strong> and matching elements <em>x<\/em><sub>0<\/sub>, <em>x<\/em><sub>1<\/sub>, &#8230;, <em>x<sub>n<\/sub><\/em> such that <em>x<\/em>=<em>x<\/em><sub>0<\/sub> \u2291<em><sub>V<\/sub><\/em><sub>1<\/sub> <em>x<\/em><sub>1<\/sub> \u2291<em><sub>V<\/sub><\/em><sub>2<\/sub> &#8230; \u2291<em><sub>V<\/sub><\/em><sub><em>n<\/em><\/sub> <em>x<sub>n<\/sub><\/em>=<em>y<\/em>.<br>The resulting family (^\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub> is another precirculation on <em>X<\/em>, and one such that <em>x<\/em> ^\u2291<em><sub>U<\/sub><\/em> <em>y<\/em> implies <em>x<\/em> \u2291<em><sub>U<\/sub><\/em> <em>y<\/em> (take the open cover consisting of the sole set <em>U<\/em>).  Let me call the new prestream (<em>X<\/em>, (^\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>) the <em>one-step cosheafification<\/em> of (<em>X<\/em>, (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>).<\/li><li>The one-step cosheafification of (<em>X<\/em>, (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>) may fail to be a stream, so we iterate the process&#8230; transfinitely.  The process stops at some ordinal, and what we obtain is the cosheafification of (<em>X<\/em>, (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>).  Put in a simpler, albeit more abstract way, there is a complete lattice of precirculations on <em>X<\/em>, and one-step cosheafification defines a monotonic, deflationary operator on that complete lattice; cosheafification finds the largest fixed point of that operator below the initial precirculation \u2291<em><sub>U<\/sub><\/em>.<\/li><\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">The lattice-theoretic view that we have seen earlier makes this easier to grasp, perhaps.  We start with a (modified) precirculation <em>circ<\/em>, namely a monotonic map <em>circ<\/em> from <strong>O<\/strong><em>X<\/em> to <strong>Pre<\/strong>(<em>X<\/em>), and below <strong>d<\/strong><em><sub>X<\/sub>.<\/em>  The cosheafification is obtained by replacing <em>circ<\/em> by the largest supremum preserving map below <em>circ<\/em>.  This can also be obtained by as the largest fixed point of some one-step cosheafification process.  It is easier to realize that there is always a largest supremum preserving map below any map <em>f<\/em>, which one obtains as the pointwise supremum of the family of all supremum preserving maps below <em>f<\/em>.  Indeed, any pointwise supremum of supremum preserving maps is itself supremum preserving.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">As in the care of any coflective subcategory, limits in <strong>Str<\/strong> are obtained by applying the coreflection functor (cosheafification) to limits computed in the larger category <strong>PreStr<\/strong>.  A nice things is that, in the case of products, cosheafification stabilizes at step 1: in other words, the product of two streams in <strong>Str<\/strong> is the <em>one-step<\/em> cosheafification of their product in <strong>PreStr<\/strong> [4, Proposition 12].<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Sheaves<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Why is the cosheafification called <em>cosheafification<\/em>?  Well, simply because streams are cosheaves of preordered sets.  In other words, the circulation condition really means that, for a prestream (<em>X<\/em>, (\u2291<em><sub>U<\/sub><\/em>)<sub><em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em><\/sub>) to be a stream, each preordered set (<em>U<\/em>, \u2291<em><sub>U<\/sub><\/em>) must be the colimit, taken in the category of preordered sets and monotonic maps, of the diagram of preordered sets (<em>V<\/em>, \u2291<em><sub>V<\/sub><\/em>), where <em>V<\/em> ranges over the open subsets of <em>U<\/em>, and with the obvious inclusions as morphisms.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">But talking about (co)sheaves would lead me too far away for now, so I&#8217;ll talk about them another time!<\/p>\n\n\n\n<ol class=\"wp-block-list\"><li>Marco Grandis.  <em><a href=\"https:\/\/www.cambridge.org\/fr\/academic\/subjects\/mathematics\/geometry-and-topology\/directed-algebraic-topology-models-non-reversible-worlds?format=HB&amp;isbn=9780521760362\">Directed Algebraic Topology: Models of Non-Reversible Worlds<\/a><\/em>. New Mathematical Monographs <strong>13<\/strong>, Cambridge University Press, 2009.  ISBN-13 978-0-521-76036-2.<\/li><li>Sanjeevi Krishnan. <a href=\"https:\/\/link.springer.com\/article\/10.1007\/s10485-008-9140-9\">A convenient category of locally preordered spaces<\/a>.&nbsp;<a href=\"https:\/\/www.springer.com\/journal\/10485\">Applied Categorical Structures,<\/a> 17(5):445\u2013466, 2009.<\/li><li>Emmanuel Haucourt.  <a href=\"https:\/\/www.sciencedirect.com\/science\/article\/pii\/S1571066112000096\">Streams, d-spaces, and their fundamental categories<\/a>.  <a href=\"https:\/\/www.sciencedirect.com\/science\/article\/pii\/S1571066112000096\">Electronic Notes in Theoretical Computer Science<\/a>&nbsp;<a href=\"https:\/\/www.sciencedirect.com\/science\/journal\/15710661\/283\/supp\/C\">283<\/a>, pages 111\u2013151, 2012.<\/li><li>Jean Goubault-Larrecq.  <a href=\"https:\/\/hal.inria.fr\/hal-03189894\/\">Exponentiable streams and prestreams<\/a>.  <a href=\"https:\/\/link.springer.com\/journal\/10485\"><em>Applied Categorical Structures<\/em><\/a>&nbsp;22,&nbsp;pages&nbsp;515\u2013549, 2014.  The published version, available from <a href=\"https:\/\/link.springer.com\/article\/10.1007\/s10485-013-9315-x\">Springer Link<\/a>, contains two mistakes, which are repaired in the <a href=\"https:\/\/hal.inria.fr\/hal-03189894\/\">HAL report<\/a>.<\/li><\/ol>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"alignright is-resized\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2016\/08\/jgl-2011.png\" alt=\"jgl-2011\" class=\"wp-image-993\" width=\"60\" height=\"83\"\/><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">\u2014 <a href=\"https:\/\/www.lsv.ens-paris-saclay.fr\/~goubault\/?l=en\">Jean Goubault-Larrecq<\/a> (August 24th, 2021)<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Last time, I had introduced two models of directed topological spaces, namely of topological spaces with a local direction of time: Marco Grandis&#8217; d-spaces [1], which are pairs (X, dX) of a topological space and of a collection dX of &hellip; <a href=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/?page_id=4067\">Continue reading <span class=\"meta-nav\">&rarr;<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"_crdt_document":"","footnotes":""},"class_list":["post-4067","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=\/wp\/v2\/pages\/4067","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=4067"}],"version-history":[{"count":50,"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=\/wp\/v2\/pages\/4067\/revisions"}],"predecessor-version":[{"id":5343,"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=\/wp\/v2\/pages\/4067\/revisions\/5343"}],"wp:attachment":[{"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4067"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}