-------- Dridi Boukelmoune writes: > On Tue, Oct 7, 2025 at 10:56=E2=80=AFAM Poul-Henning Kamp <[email protected].= > dk> wrote: > > > > As tempted as I am to go "I have UTF-8 and I'm not afraid to use > > it" and rename to "=C3=A9cluse", I personally see it more as some sort > > of distribution mechanism than as a flow-control mechanism. > > I talked about coordination, not flow control. The principle is two > parts of a canal operating at different levels (client vs backend) and > the sluice coordinating the transit of boats (body bytes) from one > level to the other. Sure that involve some amounts of flow control, > but that's just means to an end.
/me inhales deeply... The reason you need the sluice in the first place is because you want to control the flow of the river with a dam :-) But the reason I say it looks more like a "distrbution mechanism" is that in difference from a sluice, which has one entrace and one exit, we have one entrance and as many exits as there are clients streaming this object. If you want a precise technological prior art, it would be the "fan-out" points in private telex networks. For instance a news-providers like Reuters, would have one leased telex-circuit from London to Paris, where it would be fanned out to Nice, Strassbourg, Bruxelles (and all the local Paris newspapers), and from Nice to Torino, and from there to Milano and Rome etc. forming a tree structure which a minmal (in terms of line costs) spanning tree. In North America FCC made a rule that you paid AT&T the cost of the minimal spanning tree, which is why Knuth's algorithm became so important so fast. Both customers and AT&T gamed that FCC rule. Companies cost-optimized their nets by adding nodes they did not need in remote hamlets like Hoople, North Dakota, because gave them a cheaper connection from Chicago to San Francisco etc. AT&T on the other hand implemented the actual network the way which made most economic sense to them, which meant that it was anyones guess what the delay in the total network would be. That again got reflected in IBM's VTAM which has some very bizarre configuration options for node timeouts and ... Sorry, we were taking about what again ? :-) If you want a really weird, but precise name suggestion: "Hopper" That word searches horribly but try "input hopper" or "grain hopper" and try to generalize yourself. Poul-Henning -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 [email protected] | TCP/IP since RFC 956 FreeBSD committer | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence. _______________________________________________ varnish-dev mailing list [email protected] https://www.varnish-cache.org/lists/mailman/listinfo/varnish-dev
