<!DOCTYPE html>
<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body>
    When I introduced the VGV acronym, I used Gaming as the G, but that
    was probably misleading.  There are many "use cases" other than
    gaming.   A better term might be Interactive, and a VIV acronym.  <br>
    <br>
    Personally, I've been experimenting with "home automation", and have
    found the cloud-based schemes to be intolerable.   When you flip a
    light switch, you expect the light to come on almost instantly, and
    cloud-based mechanisms are disturbingly inconsistent, possibly due
    to variability in latency.   Lights turning on 20 seconds after you
    flip the switch is unacceptable. <br>
    <br>
    There are other such "serious" Interactive use cases not related to
    Gaming.  Consider, for example, a "driverless" vehicle which relies
    on the ability for a human at some remote control center to "take
    over" control of a vehicle when necessary.   Or a power plant, or
    medical device, or ...<br>
    <br>
    IMHO, "Use Cases" are the foundation to setting good policy and
    regulation.  What should Users expect to be able to do when using
    the Internet?   <br>
    <br>
    Back in the early days, the design of Internet mechanisms was driven
    by several "use cases".  Since the project was funded by the
    military, the use cases were also military.   But humans need
    similar kinds of communications, whether they are exchanging
    commands to the troops or tweets to the forums.<br>
    <br>
    One such "use case" motivated the concern for latency.  The scenario
    was simple.  An army is engaged in some action, and an online
    conference is being held amongst all the players, from the generals
    in HQs to the field commanders in tents or even jeeps or airplanes. 
    The conference supports multimedia interactions, using something
    like a "shared whiteboard" that everyone can see (video was just a
    dream in 1980) and change using a mouse or some kind of pointer
    device.<br>
    <br>
    In such a scenario, the participants might be viewing a map, and
    exchanging information while pointing on the map.  "The enemy HQ is
    here."  "Our patrol is here."  "Send the battalion here."  "We'll
    order artillery to strike here."  It's very important that the
    pointing is synchronized with the speech, and that the graphics and
    speech are delivered intact.<br>
    <br>
    Those scenarios motivated the inclusion of Internet mechanisms such
    as SQ, TTL, TOS et al.<br>
    <tt> </tt><br>
    Such scenarios are used today, quite frequently, in normal everyday
    life by humans in all sorts of activities.  They're no longer just
    military situations.<br>
    <br>
    Considering "freedom of speech", what are the Use Cases that the
    Internet must support?  Setting aside the question of what can be
    said, are there other aspects of freedom of speech?  <br>
    <br>
    One example - assuming people should have the right to speak
    anonymously, should they also have the right to speak
    non-anonymously?  Should someone have the right to know that speech
    attributed to some person actually was spoken by that person?<br>
    <br>
    Today, regulators are dealing with this question in other venues,
    e.g., by adding, and enforcing, the notion of "verified" telephone
    numbers, in response to the tsunami of robocalls, phishing, and such
    activities.  Should the Internet also provide a way to verify the
    sources of emails, blog posts, etc.?<br>
    <br>
    I've "digitally signed" this message, using the technology which I
    found in my email program.   Maybe that will enable you to believe
    that I actually wrote this message.   But I rarely receive any
    "signed" email, except occasionally from members of the Technorati. 
    Perhaps the requirement to be anonymous is just part of the Internet
    now?   Or is that a Use Case that simply hasn't been developed,
    propagated, and enforced yet?<br>
    <br>
    What are the "Use Cases" that the Internet must support, and that
    regulators should enforce?<br>
    <br>
    Jack Haverty<br>
     <br>
    <br>
    <br>
    <div class="moz-cite-prefix">On 10/17/23 10:26, Spencer Sevilla via
      Nnagain wrote:<br>
    </div>
    <blockquote type="cite"
      cite="mid:BFCFC49C-24D2-4243-9602-596A7599DB64@gmail.com">
      <meta http-equiv="content-type" content="text/html; charset=UTF-8">
      I know this is a small side note but I felt compelled to speak up
      in defense of online gaming. I’m not a gamer at all and up till a
      year or two ago, would have agreed with Dick’s take about benefit
      to “society as a whole.” However, lately I’ve started hearing some
      research on the benefits of groups of friends using online games
      to socialize together, effectively using the game primarily as a
      group call.
      <div><br>
      </div>
      <div>There’s also this project, where people have collected
        banned/censored books into a library in Minecraft. Specifically
        as a solution to contexts where regulators/censors ban and
        monitor content through other channels (websites etc) but don’t
        surveil Minecraft... Presumably because they share Dick’s
        opinion ;-) <a href="https://www.uncensoredlibrary.com/en"
          moz-do-not-send="true" class="moz-txt-link-freetext">https://www.uncensoredlibrary.com/en</a>
        <div>
          <div><br>
            <blockquote type="cite">
              <div>On Oct 17, 2023, at 03:26, Sebastian Moeller via
                Nnagain <a class="moz-txt-link-rfc2396E" href="mailto:nnagain@lists.bufferbloat.net"><nnagain@lists.bufferbloat.net></a> wrote:</div>
              <br class="Apple-interchange-newline">
              <div>
                <meta charset="UTF-8">
                <span
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;">Hi
                  Richard,</span><br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <blockquote type="cite"
style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">On
                  Oct 16, 2023, at 20:04, Dick Roy
                  <a class="moz-txt-link-rfc2396E" href="mailto:dickroy@alum.mit.edu"><dickroy@alum.mit.edu></a> wrote:<br>
                  <br>
                  Good points all, Sebastien.  How to "trade-off" a
                  fixed capacity amongst many users is ultimately a game
                  theoretic problem when users are allowed to make
                  choices, which is certainly the case here.  Secondly,
                  any network that can and does generate "more traffic"
                  (aka overhead such as ACKs NACKs and retries) reduces
                  the capacity of the network, and ultimately can lead
                  to the "user" capacity going to zero!  Such is life in
                  the fast lane (aka the internet).<br>
                  <br>
                  Lastly, on the issue of low-latency real-time
                  experience, there are many applications that need/want
                  such capabilities that actually have a net benefit to
                  the individuals involved AND to society as a whole.
                   IMO, interactive gaming is NOT one of those.<br>
                </blockquote>
                <br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <span class="Apple-tab-span"
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: pre; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">       </span><span
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;">[SM]
                  Yes, gaming is one obvious example of a class of uses
                  that work best with low latency and low jitter, not
                  necessarily an example for a use-case worthy enough to
                  justify the work required to increase the
                  responsiveness of the internet. Other examples are
                  video conferences, VoIP, in extension of both musical
                  collaboration over the internet, and surprising to
                  some even plain old web browsing (it often needs to
                  first read a page before being able to follow links
                  and load resources, and every read takes at best a
                  single RTT). None of these are inherently beneficial
                  or detrimental to individuals or society, but most can
                  be used to improve the status quo... I would argue
                  that in the last 4 years the relevance of interactive
                  use-cases has been made quite clear to a lot of
                  folks...</span><br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <blockquote type="cite"
style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">OK,
                  so now you know I don't engage in these time sinks
                  with no redeeming social value.:)<br>
                </blockquote>
                <br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <span class="Apple-tab-span"
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: pre; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">       </span><span
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;">[SM]
                  Duly noted ;)</span><br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <blockquote type="cite"
style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">Since
                  it is not hard to argue that just like power
                  distribution, information exchange/dissemination is
                  "in the public interest", the question becomes "Do we
                  allow any and all forms of information
                  exchange/dissemination over what is becoming something
                  akin to a public utility?"  FWIW, I don't know the
                  answer to this question! :)<br>
                </blockquote>
                <br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <span class="Apple-tab-span"
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: pre; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">       </span><span
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;">[SM]
                  This is an interesting question and one (only)
                  tangentially related to network neutrality... it is
                  more related to freedom of speech and limits thereof.
                  Maybe a question for another mailing list? Certainly
                  one meriting a topic change...</span><br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <span
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;">Regards</span><br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <span class="Apple-tab-span"
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: pre; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">       </span><span
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;">Sebastian</span><br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <blockquote type="cite"
style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;"><br>
                  Cheers,<br>
                  <br>
                  RR<br>
                  <br>
                  -----Original Message-----<br>
                  From: Sebastian Moeller [<a class="moz-txt-link-freetext" href="mailto:moeller0@gmx.de">mailto:moeller0@gmx.de</a>]<span
                    class="Apple-converted-space"> </span><br>
                  Sent: Monday, October 16, 2023 10:36 AM<br>
                  To: <a class="moz-txt-link-abbreviated" href="mailto:dickroy@alum.mit.edu">dickroy@alum.mit.edu</a>; Network Neutrality is back!
                  Let´s make the technical aspects heard this time!<br>
                  Subject: Re: [NNagain] transit and peering costs
                  projections<br>
                  <br>
                  Hi Richard,<br>
                  <br>
                  <br>
                  <blockquote type="cite">On Oct 16, 2023, at 19:01,
                    Dick Roy via Nnagain
                    <a class="moz-txt-link-rfc2396E" href="mailto:nnagain@lists.bufferbloat.net"><nnagain@lists.bufferbloat.net></a> wrote:<br>
                    <br>
                    Just an observation:  ANY type of congestion control
                    that changes application behavior in response to
                    congestion, or predicted congestion (ENC), begs the
                    question "How does throttling of application
                    information exchange rate (aka behavior) affect the
                    user experience and will the user tolerate it?"<span
                      class="Apple-converted-space"> </span><br>
                  </blockquote>
                  <br>
                  <span class="Apple-tab-span" style="white-space: pre;">     </span>[SM]
                  The trade-off here is, if the application does not
                  respond (or rather if no application would respond) we
                  would end up with congestion collapse where no
                  application would gain much of anything as the network
                  busies itself trying to re-transmit dropped packets
                  without making much head way... Simplistic game theory
                  application might imply that individual applications
                  could try to game this, and generally that seems to be
                  true, but we have remedies for that available..<br>
                  <br>
                  <br>
                  <blockquote type="cite"><br>
                    Given any (complex and packet-switched) network
                    topology of interconnected nodes and links, each
                    with possible a different capacity and
                    characteristics, such as the internet today, IMO the
                    two fundamental questions are:<br>
                    <br>
                    1) How can a given network be operated/configured so
                    as to maximize aggregate throughput (i.e. achieve
                    its theoretical capacity), and<br>
                    2) What things in the network need to change to
                    increase the throughput (aka parameters in the
                    network with the largest Lagrange multipliers
                    associated with them)?<br>
                  </blockquote>
                  <br>
                  <span class="Apple-tab-span" style="white-space: pre;">     </span>[SM]
                  The thing is we generally know how to maximize
                  (average) throughput, just add (over-)generous amounts
                  of buffering, the problem is that this screws up the
                  other important quality axis, latency... We ideally
                  want low latency and even more low latency variance
                  (aka jitter) AND high throughput... Turns out though
                  that above a certain throughput threshold* many users
                  do not seem to care all that much for more throughput
                  as long as interactive use cases are sufficiently
                  responsive... but high responsiveness requires low
                  latency and low jitter... This is actually a good
                  thing, as that means we do not necessarily aim for
                  100% utilization (almost requiring deep buffers and
                  hence resulting in compromised latency) but can get
                  away with say 80-90% where shallow buffers will do (or
                  rather where buffer filling stays shallow, there is
                  IMHO still value in having deep buffers for rare
                  events that need it).<br>
                  <br>
                  <br>
                  <br>
                  *) This is not a hard physical law so the exact
                  threshold is not set in stone, but unless one has many
                  parallel users, something in the 20-50 Mbps range is
                  plenty and that is only needed in the "loaded"
                  direction, that is for pure consumers the upload can
                  be thinner, for pure producers the download can be
                  thinner.<br>
                  <br>
                  <br>
                  <blockquote type="cite"><br>
                    I am not an expert in this field,<br>
                  </blockquote>
                  <br>
                       [SM] Nor am I, I come from the wet-ware side of
                  things so not even soft- or hard-ware ;)<br>
                  <br>
                  <br>
                  <blockquote type="cite">however it seems to me that
                    answers to these questions would be useful, assuming
                    they are not yet available!<br>
                    <br>
                    Cheers,<br>
                    <br>
                    RR<br>
                    <br>
                    <br>
                    <br>
                    -----Original Message-----<br>
                    From: Nnagain
                    [<a class="moz-txt-link-freetext" href="mailto:nnagain-bounces@lists.bufferbloat.net">mailto:nnagain-bounces@lists.bufferbloat.net</a>] On
                    Behalf Of rjmcmahon via Nnagain<br>
                    Sent: Sunday, October 15, 2023 1:39 PM<br>
                    To: Network Neutrality is back! Let´s make the
                    technical aspects heard this time!<br>
                    Cc: rjmcmahon<br>
                    Subject: Re: [NNagain] transit and peering costs
                    projections<br>
                    <br>
                    Hi Jack,<br>
                    <br>
                    Thanks again for sharing. It's very interesting to
                    me.<br>
                    <br>
                    Today, the networks are shifting from capacity
                    constrained to latency<span
                      class="Apple-converted-space"> </span><br>
                    constrained, as can be seen in the IX discussions
                    about how the speed of<span
                      class="Apple-converted-space"> </span><br>
                    light over fiber is too slow even between Houston
                    & Dallas.<br>
                    <br>
                    The mitigations against standing queues (which cause
                    bloat today) are:<br>
                    <br>
                    o) Shrink the e2e bottleneck queue so it will drop
                    packets in a flow and<span
                      class="Apple-converted-space"> </span><br>
                    TCP will respond to that "signal"<br>
                    o) Use some form of ECN marking where the network
                    forwarding plane<span class="Apple-converted-space"> </span><br>
                    ultimately informs the TCP source state machine so
                    it can slow down or<span
                      class="Apple-converted-space"> </span><br>
                    pace effectively. This can be an earlier feedback
                    signal and, if done<span
                      class="Apple-converted-space"> </span><br>
                    well, can inform the sources to avoid bottleneck
                    queuing. There are<span
                      class="Apple-converted-space"> </span><br>
                    couple of approaches with ECN. Comcast is trialing
                    L4S now which seems<span
                      class="Apple-converted-space"> </span><br>
                    interesting to me as a WiFi test & measurement
                    engineer. The jury is<span
                      class="Apple-converted-space"> </span><br>
                    still out on this and measurements are needed.<br>
                    o) Mitigate source side bloat via TCP_NOTSENT_LOWAT<br>
                    <br>
                    The QoS priority approach per congestion is
                    orthogonal by my judgment as<span
                      class="Apple-converted-space"> </span><br>
                    it's typically not supported e2e, many networks will
                    bleach DSCP<span class="Apple-converted-space"> </span><br>
                    markings. And it's really too late by my judgment.<br>
                    <br>
                    Also, on clock sync, yes your generation did us both
                    a service and<span class="Apple-converted-space"> </span><br>
                    disservice by getting rid of the PSTN TDM clock ;)
                    So IP networking<span class="Apple-converted-space"> </span><br>
                    devices kinda ignored clock sync, which makes e2e
                    one way delay (OWD)<span
                      class="Apple-converted-space"> </span><br>
                    measurements impossible. Thankfully, the GPS atomic
                    clock is now<span class="Apple-converted-space"> </span><br>
                    available mostly everywhere and many devices use
                    TCXO oscillators so<span
                      class="Apple-converted-space"> </span><br>
                    it's possible to get clock sync and use oscillators
                    that can minimize<span class="Apple-converted-space"> </span><br>
                    drift. I pay $14 for a Rpi4 GPS chip with pulse per
                    second as an<span class="Apple-converted-space"> </span><br>
                    example.<br>
                    <br>
                    It seems silly to me that clocks aren't synced to
                    the GPS atomic clock<span
                      class="Apple-converted-space"> </span><br>
                    even if by a proxy even if only for measurement and
                    monitoring.<br>
                    <br>
                    Note: As Richard Roy will point out, there really is
                    no such thing as<span class="Apple-converted-space"> </span><br>
                    synchronized clocks across geographies per general
                    relativity - so those<span
                      class="Apple-converted-space"> </span><br>
                    syncing clocks need to keep those effects in mind. I
                    limited the iperf 2<span
                      class="Apple-converted-space"> </span><br>
                    timestamps to microsecond precision in hopes
                    avoiding those issues.<br>
                    <br>
                    Note: With WiFi, a packet drop can occur because an
                    intermittent RF<span class="Apple-converted-space"> </span><br>
                    channel condition. TCP can't tell the difference
                    between an RF drop vs a<span
                      class="Apple-converted-space"> </span><br>
                    congested queue drop. That's another reason ECN
                    markings from network<span
                      class="Apple-converted-space"> </span><br>
                    devices may be better than dropped packets.<br>
                    <br>
                    Note: I've added some iperf 2 test support around
                    pacing as that seems<span
                      class="Apple-converted-space"> </span><br>
                    to be the direction the industry is heading as
                    networks are less and<span
                      class="Apple-converted-space"> </span><br>
                    less capacity strained and user quality of
                    experience is being driven by<span
                      class="Apple-converted-space"> </span><br>
                    tail latencies. One can also test with the Prague
                    CCA for the L4S<span class="Apple-converted-space"> </span><br>
                    scenarios. (This is a fun project:
                    <a class="moz-txt-link-freetext" href="https://www.l4sgear.com/">https://www.l4sgear.com/</a> and fairly<span
                      class="Apple-converted-space"> </span><br>
                    low cost)<br>
                    <br>
                    --fq-rate n[kmgKMG]<br>
                    Set a rate to be used with fair-queuing based
                    socket-level pacing, in<span
                      class="Apple-converted-space"> </span><br>
                    bytes or bits per second. Only available on
                    platforms supporting the<span
                      class="Apple-converted-space"> </span><br>
                    SO_MAX_PACING_RATE socket option. (Note: Here the
                    suffixes indicate<span class="Apple-converted-space"> </span><br>
                    bytes/sec or bits/sec per use of uppercase or
                    lowercase, respectively)<br>
                    <br>
                    --fq-rate-step n[kmgKMG]<br>
                    Set a step of rate to be used with fair-queuing
                    based socket-level<span
                      class="Apple-converted-space"> </span><br>
                    pacing, in bytes or bits per second. Step occurs
                    every<span class="Apple-converted-space"> </span><br>
                    fq-rate-step-interval (defaults to one second)<br>
                    <br>
                    --fq-rate-step-interval n<br>
                    Time in seconds before stepping the fq-rate<br>
                    <br>
                    Bob<br>
                    <br>
                    PS. Iperf 2 man page
                    <a class="moz-txt-link-freetext" href="https://iperf2.sourceforge.io/iperf-manpage.html">https://iperf2.sourceforge.io/iperf-manpage.html</a><br>
                    <br>
                    <blockquote type="cite">The "VGV User" (Voice,
                      Gaming, Videoconferencing) cares a lot about<br>
                      latency.   It's not just "rewarding" to have lower
                      latencies; high<br>
                      latencies may make VGV unusable.   Average (or
                      "typical") latency as<br>
                      the FCC label proposes isn't a good metric to
                      judge usability.  A path<br>
                      which has high variance in latency can be unusable
                      even if the average<br>
                      is quite low.   Having your voice or video or
                      gameplay "break up"<br>
                      every minute or so when latency spikes to 500 msec
                      makes the "user<br>
                      experience" intolerable.<br>
                      <br>
                      A few years ago, I ran some simple "ping" tests to
                      help a friend who<br>
                      was trying to use a gaming app.  My data was only
                      for one specific<br>
                      path so it's anecdotal.  What I saw was surprising
                      - zero data loss,<br>
                      every datagram was delivered, but occasionally a
                      datagram would take<br>
                      up to 30 seconds to arrive.  I didn't have the
                      ability to poke around<br>
                      inside, but I suspected it was an experience of
                      "bufferbloat", enabled<br>
                      by the dramatic drop in price of memory over the
                      decades.<br>
                      <br>
                      It's been a long time since I was involved in
                      operating any part of<br>
                      the Internet, so I don't know much about the inner
                      workings today.<br>
                      Apologies for my ignorance....<br>
                      <br>
                      There was a scenario in the early days of the
                      Internet for which we<br>
                      struggled to find a technical solution.  Imagine
                      some node in the<br>
                      bowels of the network, with 3 connected "circuits"
                      to some other<br>
                      nodes.  On two of those inputs, traffic is
                      arriving to be forwarded<br>
                      out the third circuit.  The incoming flows are
                      significantly more than<br>
                      the outgoing path can accept.<br>
                      <br>
                      What happens?   How is "backpressure" generated so
                      that the incoming<br>
                      flows are reduced to the point that the outgoing
                      circuit can handle<br>
                      the traffic?<br>
                      <br>
                      About 45 years ago, while we were defining TCPV4,
                      we struggled with<br>
                      this issue, but didn't find any consensus
                      solutions.  So "placeholder"<br>
                      mechanisms were defined in TCPV4, to be replaced
                      as research continued<br>
                      and found a good solution.<br>
                      <br>
                      In that "placeholder" scheme, the "Source Quench"
                      (SQ) IP message was<br>
                      defined; it was to be sent by a switching node
                      back toward the sender<br>
                      of any datagram that had to be discarded because
                      there wasn't any<br>
                      place to put it.<br>
                      <br>
                      In addition, the TOS (Type Of Service) and TTL
                      (Time To Live) fields<br>
                      were defined in IP.<br>
                      <br>
                      TOS would allow the sender to distinguish
                      datagrams based on their<br>
                      needs.  For example, we thought "Interactive"
                      service might be needed<br>
                      for VGV traffic, where timeliness of delivery was
                      most important.<span class="Apple-converted-space"> </span><br>
                      "Bulk" service might be useful for activities like
                      file transfers,<br>
                      backups, et al.   "Normal" service might now mean
                      activities like<br>
                      using the Web.<br>
                      <br>
                      The TTL field was an attempt to inform each
                      switching node about the<br>
                      "expiration date" for a datagram.   If a node
                      somehow knew that a<br>
                      particular datagram was unlikely to reach its
                      destination in time to<br>
                      be useful (such as a video datagram for a frame
                      that has already been<br>
                      displayed), the node could, and should, discard
                      that datagram to free<br>
                      up resources for useful traffic.  Sadly we had no
                      mechanisms for<br>
                      measuring delay, either in transit or in queuing,
                      so TTL was defined<br>
                      in terms of "hops", which is not an accurate proxy
                      for time.   But<br>
                      it's all we had.<br>
                      <br>
                      Part of the complexity was that the "flow control"
                      mechanism of the<br>
                      Internet had put much of the mechanism in the
                      users' computers' TCP<br>
                      implementations, rather than the switches which
                      handle only IP.<br>
                      Without mechanisms in the users' computers, all a
                      switch could do is<br>
                      order more circuits, and add more memory to the
                      switches for queuing.<span
                        class="Apple-converted-space"> </span><br>
                      Perhaps that led to "bufferbloat".<br>
                      <br>
                      So TOS, SQ, and TTL were all placeholders, for
                      some mechanism in a<br>
                      future release that would introduce a "real" form
                      of Backpressure and<br>
                      the ability to handle different types of traffic.
                        Meanwhile, these<br>
                      rudimentary mechanisms would provide some flow
                      control. Hopefully the<br>
                      users' computers sending the flows would respond
                      to the SQ<br>
                      backpressure, and switches would prioritize
                      traffic using the TTL and<br>
                      TOS information.<br>
                      <br>
                      But, being way out of touch, I don't know what
                      actually happens<br>
                      today.  Perhaps the current operators and current
                      government watchers<br>
                      can answer?:git clone
                      <a class="moz-txt-link-freetext" href="https://rjmcmahon@git.code.sf.net/p/iperf2/code">https://rjmcmahon@git.code.sf.net/p/iperf2/code</a><span
                        class="Apple-converted-space"> </span><br>
                      iperf2-code<br>
                      <br>
                      1/ How do current switches exert Backpressure to
                       reduce competing<br>
                      traffic flows?  Do they still send SQs?<br>
                      <br>
                      2/ How do the current and proposed government
                      regulations treat the<br>
                      different needs of different types of traffic,
                      e.g., "Bulk" versus<br>
                      "Interactive" versus "Normal"?  Are Internet
                      carriers permitted to<br>
                      treat traffic types differently?  Are they
                      permitted to charge<br>
                      different amounts for different types of service?<br>
                      <br>
                      Jack Haverty<br>
                      <br>
                      On 10/15/23 09:45, Dave Taht via Nnagain wrote:<br>
                      <blockquote type="cite">For starters I would like
                        to apologize for cc-ing both nanog and my<br>
                        new nn list. (I will add sender filters)<br>
                        <br>
                        A bit more below.<br>
                        <br>
                        On Sun, Oct 15, 2023 at 9:32 AM Tom Beecher
                        <a class="moz-txt-link-rfc2396E" href="mailto:beecher@beecher.cc"><beecher@beecher.cc></a><span
                          class="Apple-converted-space"> </span><br>
                        wrote:<br>
                        <blockquote type="cite">
                          <blockquote type="cite">So for now, we'll keep
                            paying for transit to get to the others<span
                              class="Apple-converted-space"> </span><br>
                            (since it’s about as much as transporting
                            IXP from Dallas), and<span
                              class="Apple-converted-space"> </span><br>
                            hoping someone at Google finally sees
                            Houston as more than a third<span
                              class="Apple-converted-space"> </span><br>
                            rate city hanging off of Dallas. Or… someone
                            finally brings a<span
                              class="Apple-converted-space"> </span><br>
                            worthwhile IX to Houston that gets us more
                            than peering to Kansas<span
                              class="Apple-converted-space"> </span><br>
                            City. Yeah, I think the former is more
                            likely. 😊<br>
                          </blockquote>
                          <br>
                          There is often a chicken/egg scenario here
                          with the economics. As an<span
                            class="Apple-converted-space"> </span><br>
                          eyeball network, your costs to build out and
                          connect to Dallas are<span
                            class="Apple-converted-space"> </span><br>
                          greater than your transit cost, so you do
                          that. Totally fair.<br>
                          <br>
                          However think about it from the content side.
                          Say I want to build<span
                            class="Apple-converted-space"> </span><br>
                          into to Houston. I have to put routers in, and
                          a bunch of cache<span
                            class="Apple-converted-space"> </span><br>
                          servers, so I have capital outlay , plus opex
                          for space, power,<span
                            class="Apple-converted-space"> </span><br>
                          IX/backhaul/transit costs. That's not cheap,
                          so there's a lot of<span
                            class="Apple-converted-space"> </span><br>
                          calculations that go into it. Is there enough
                          total eyeball traffic<span
                            class="Apple-converted-space"> </span><br>
                          there to make it worth it? Is saving 8-10ms
                          enough of a performance<span
                            class="Apple-converted-space"> </span><br>
                          boost to justify the spend? What are the long
                          term trends in that<span
                            class="Apple-converted-space"> </span><br>
                          market? These answers are of course different
                          for a company running<span
                            class="Apple-converted-space"> </span><br>
                          their own CDN vs the commercial CDNs.<br>
                          <br>
                          I don't work for Google and obviously don't
                          speak for them, but I<span
                            class="Apple-converted-space"> </span><br>
                          would suspect that they're happy to eat a
                          8-10ms performance hit to<span
                            class="Apple-converted-space"> </span><br>
                          serve from Dallas , versus the amount of
                          capital outlay to build out<span
                            class="Apple-converted-space"> </span><br>
                          there right now.<br>
                        </blockquote>
                        The three forms of traffic I care most about are
                        voip, gaming, and<br>
                        videoconferencing, which are rewarding to have
                        at lower latencies.<br>
                        When I was a kid, we had switched phone
                        networks, and while the sound<br>
                        quality was poorer than today, the voice latency
                        cross-town was just<br>
                        like "being there". Nowadays we see 500+ms
                        latencies for this kind of<br>
                        traffic.<br>
                        <br>
                        As to how to make calls across town work that
                        well again, cost-wise, I<br>
                        do not know, but the volume of traffic that
                        would be better served by<br>
                        these interconnects quite low, respective to the
                        overall gains in<br>
                        lower latency experiences for them.<br>
                        <br>
                        <br>
                        <br>
                        <blockquote type="cite">On Sat, Oct 14, 2023 at
                          11:47 PM Tim Burke <a class="moz-txt-link-rfc2396E" href="mailto:tim@mid.net"><tim@mid.net></a> wrote:<br>
                          <blockquote type="cite">I would say that a
                            1Gbit IP transit in a carrier neutral DC can
                            be<span class="Apple-converted-space"> </span><br>
                            had for a good bit less than $900 on the
                            wholesale market.<br>
                            <br>
                            Sadly, IXP’s are seemingly turning into a
                            pay to play game, with<span
                              class="Apple-converted-space"> </span><br>
                            rates almost costing as much as transit in
                            many cases after you<span
                              class="Apple-converted-space"> </span><br>
                            factor in loop costs.<br>
                            <br>
                            For example, in the Houston market (one of
                            the largest and fastest<span
                              class="Apple-converted-space"> </span><br>
                            growing regions in the US!), we do not have
                            a major IX, so to get up<span
                              class="Apple-converted-space"> </span><br>
                            to Dallas it’s several thousand for a 100g
                            wave, plus several<span
                              class="Apple-converted-space"> </span><br>
                            thousand for a 100g port on one of those
                            major IXes. Or, a better<span
                              class="Apple-converted-space"> </span><br>
                            option, we can get a 100g flat internet
                            transit for just a little<span
                              class="Apple-converted-space"> </span><br>
                            bit more.<br>
                            <br>
                            Fortunately, for us as an eyeball network,
                            there are a good number<span
                              class="Apple-converted-space"> </span><br>
                            of major content networks that are allowing
                            for private peering in<span
                              class="Apple-converted-space"> </span><br>
                            markets like Houston for just the cost of a
                            cross connect and a QSFP<span
                              class="Apple-converted-space"> </span><br>
                            if you’re in the right DC, with Google and
                            some others being the<span
                              class="Apple-converted-space"> </span><br>
                            outliers.<br>
                            <br>
                            So for now, we'll keep paying for transit to
                            get to the others<span
                              class="Apple-converted-space"> </span><br>
                            (since it’s about as much as transporting
                            IXP from Dallas), and<span
                              class="Apple-converted-space"> </span><br>
                            hoping someone at Google finally sees
                            Houston as more than a third<span
                              class="Apple-converted-space"> </span><br>
                            rate city hanging off of Dallas. Or… someone
                            finally brings a<span
                              class="Apple-converted-space"> </span><br>
                            worthwhile IX to Houston that gets us more
                            than peering to Kansas<span
                              class="Apple-converted-space"> </span><br>
                            City. Yeah, I think the former is more
                            likely. 😊<br>
                            <br>
                            See y’all in San Diego this week,<br>
                            Tim<br>
                            <br>
                            On Oct 14, 2023, at 18:04, Dave Taht
                            <a class="moz-txt-link-rfc2396E" href="mailto:dave.taht@gmail.com"><dave.taht@gmail.com></a> wrote:<br>
                            <blockquote type="cite">This set of
                              trendlines was very interesting.
                              Unfortunately the<span
                                class="Apple-converted-space"> </span><br>
                              data<br>
                              stops in 2015. Does anyone have more
                              recent data?<br>
                              <br>
<a class="moz-txt-link-freetext" href="https://drpeering.net/white-papers/Internet-Transit-Pricing-Historical-And-Projected.php">https://drpeering.net/white-papers/Internet-Transit-Pricing-Historical-And-Projected.php</a><br>
                              <br>
                              I believe a gbit circuit that an ISP can
                              resell still runs at about<br>
                              $900 - $1.4k (?) in the usa? How about
                              elsewhere?<br>
                              <br>
                              ...<br>
                              <br>
                              I am under the impression that many IXPs
                              remain very successful,<br>
                              states without them suffer, and I also
                              find the concept of doing<span
                                class="Apple-converted-space"> </span><br>
                              micro<br>
                              IXPs at the city level, appealing, and now
                              achievable with cheap<span
                                class="Apple-converted-space"> </span><br>
                              gear.<br>
                              Finer grained cross connects between telco
                              and ISP and IXP would<span
                                class="Apple-converted-space"> </span><br>
                              lower<br>
                              latencies across town quite hugely...<br>
                              <br>
                              PS I hear ARIN is planning on dropping the
                              price for, and bundling<span
                                class="Apple-converted-space"> </span><br>
                              3<br>
                              BGP AS numbers at a time, as of the end of
                              this year, also.<br>
                              <br>
                              <br>
                              <br>
                              --<br>
                              Oct 30:<span class="Apple-converted-space"> </span><br>
<a class="moz-txt-link-freetext" href="https://netdevconf.info/0x17/news/the-maestro-and-the-music-bof.html">https://netdevconf.info/0x17/news/the-maestro-and-the-music-bof.html</a><br>
                              Dave Täht CSO, LibreQos<br>
                            </blockquote>
                          </blockquote>
                        </blockquote>
                        <br>
                        <br>
                      </blockquote>
                      <br>
                      _______________________________________________<br>
                      Nnagain mailing list<br>
                      <a class="moz-txt-link-abbreviated" href="mailto:Nnagain@lists.bufferbloat.net">Nnagain@lists.bufferbloat.net</a><br>
                      <a class="moz-txt-link-freetext" href="https://lists.bufferbloat.net/listinfo/nnagain">https://lists.bufferbloat.net/listinfo/nnagain</a><br>
                    </blockquote>
                    _______________________________________________<br>
                    Nnagain mailing list<br>
                    <a class="moz-txt-link-abbreviated" href="mailto:Nnagain@lists.bufferbloat.net">Nnagain@lists.bufferbloat.net</a><br>
                    <a class="moz-txt-link-freetext" href="https://lists.bufferbloat.net/listinfo/nnagain">https://lists.bufferbloat.net/listinfo/nnagain</a><br>
                    <br>
                    _______________________________________________<br>
                    Nnagain mailing list<br>
                    <a class="moz-txt-link-abbreviated" href="mailto:Nnagain@lists.bufferbloat.net">Nnagain@lists.bufferbloat.net</a><br>
                    <a class="moz-txt-link-freetext" href="https://lists.bufferbloat.net/listinfo/nnagain">https://lists.bufferbloat.net/listinfo/nnagain</a><br>
                  </blockquote>
                  <br>
                </blockquote>
                <br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <span
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;">_______________________________________________</span><br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <span
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;">Nnagain
                  mailing list</span><br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <span
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;"><a class="moz-txt-link-abbreviated" href="mailto:Nnagain@lists.bufferbloat.net">Nnagain@lists.bufferbloat.net</a></span><br
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none;">
                <span
style="caret-color: rgb(0, 0, 0); font-family: Helvetica; font-size: 12px; font-style: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; float: none; display: inline !important;"><a class="moz-txt-link-freetext" href="https://lists.bufferbloat.net/listinfo/nnagain">https://lists.bufferbloat.net/listinfo/nnagain</a></span></div>
            </blockquote>
          </div>
          <br>
        </div>
      </div>
      <br>
      <fieldset class="moz-mime-attachment-header"></fieldset>
      <pre class="moz-quote-pre" wrap="">_______________________________________________
Nnagain mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Nnagain@lists.bufferbloat.net">Nnagain@lists.bufferbloat.net</a>
<a class="moz-txt-link-freetext" href="https://lists.bufferbloat.net/listinfo/nnagain">https://lists.bufferbloat.net/listinfo/nnagain</a>
</pre>
    </blockquote>
    <br>
  </body>
</html>