From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from smtp80.ord1d.emailsrvr.com (smtp80.ord1d.emailsrvr.com [184.106.54.80]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by lists.bufferbloat.net (Postfix) with ESMTPS id 26FFF3CB37 for ; Mon, 2 Oct 2023 16:26:04 -0400 (EDT) X-Auth-ID: karl@auerbach.com Received: by smtp11.relay.ord1d.emailsrvr.com (Authenticated sender: karl-AT-auerbach.com) with ESMTPSA id 4B11A6018B for ; Mon, 2 Oct 2023 16:26:03 -0400 (EDT) Message-ID: <38508eb6-61f0-433a-9f0c-5ec1596ec179@cavebear.com> Date: Mon, 2 Oct 2023 13:26:02 -0700 MIME-Version: 1.0 User-Agent: Mozilla Thunderbird To: Nnagain Reply-To: karl@cavebear.com Content-Language: en-US From: Karl Auerbach Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 8bit X-Classification-ID: 66dd3f9b-733f-44df-bd18-4bf91bb24061-1-1 Subject: [NNagain] Intro - and questions X-BeenThere: nnagain@lists.bufferbloat.net X-Mailman-Version: 2.1.20 Precedence: list List-Id: =?utf-8?q?Network_Neutrality_is_back!_Let=C2=B4s_make_the_technical_aspects_heard_this_time!?= List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 02 Oct 2023 20:26:04 -0000 Hello everyone - I was surprised how fast this list populated and filled with useful content. I observed the "network neutrality" some years back and eventually stopped following it because it felt more like rocks randomly bouncing around inside a can than a focused inquiry towards a clearly articulated goal. I do not think network neutrality is impossible.  I see it as possible, but within limited, defined scopes.  And defining what NN, and defining those scopes, is is going to be difficult and there will be plenty of differing equally valid points of view. My first questions are definitional: - What do we mean by "net" in "net neutrality"?  Do we mean the carriage of IP packets in some what that is "fair", or at least not self-dealing by providers?  Or do we reach up into that realm felt by users - application behavior and performance?  (Example, is Google's actions vis-a-vis proprietary web extensions in Chrome a form of non-neutrality?  I could also understand claims that use of non-standardized media codecs or rapid deprecation of TLS digest algorithms are also forms of discrimination, although the latter does have some solid rationales behind it.) My own sense is that it has become more important to view "net neutrality" through the eyes of users rather than down at the packet switching layers some of us with grey beards like to live in. Someone raised the point about late 1800's discrimination on railroads (carrying coal or the infamous Standard Oil "rebates" from oil shipments by others.)  Low level price-to-transport arguments did not carry as much weight as the outrage of end users who realized they were paying more $$.  To my mind net neutrality is at least as much a political issue as a technical one, and if one wants to push a position that position is made stronger is the end user (who ultimately pays the bills) view is articulated and brought to bear. - And what do we mean by "neutrality"? I grew up in the era of the JC Whitney catalog of (largely bogus) car parts - through the right sequences of bolt-on parts you could turn a 50hp VW bug into a raging monster that got thousands of miles per gallon of gasoline.  I see lots of ads for bolt-on network widgets that make similar claims about improving network performance.  Are those things "non neutral" or do we leave them outside in a bucket labeled "worthless snake oil"? Back when we were developing entertainment grade video distribution (1995-early 2000s) Fred Baker and I worked on the RSVP (resource reservation) protocol - he did the router side, I did the client side.  At the time I always felt that making resource reservations along a network path was a good idea, but I simultaneously wondered about the costs and the opinions of those who lived in the classic "best effort" world.  And I was concerned with the ability of providers to juggle all of these things without going technically bonkers or being financially drained. I still have that concern about "who controls" and "who pays". And I'm sympathetic  to the view that "those who want better should pay more", at least to the extent that that paying more reflects actual costs of delivering more.  (There's a giant bag of worms in that phrase "actual costs", particularly in our world of multiple carriers along a path.) - How do we measure things? I have been appalled at the widespread use of "speed test" tools as the basis for product advertising and regulatory impositions. Network performance, as perceived by a user is the result of an extremely complex interactions of factors beyond mere average data delivery bytes-per-second.  It includes everything from path MTU to delay variation (jitter) to loss and reordering to burst characteristics.  It also includes ancillary stuff like DNS lookup times and even (with IPv4), ARP re-resolution time.  And as a person who mixed Sun workstations and PC's back in "the early days", I've seen how even seemingly trivial parameters (in that case, Ethernet collision backoff times) could end up giving the entire cable to Suns and lock out the PC's entirely. As some of you know, I have an evil side, an evil Mr. Hyde to my nicer Dr. Jekyll - I'm also a lawyer (California and US Federal.) So I'm kinda used to the difficult question of figuring out the line between "light grey" and "dark grey" - situations in which there is no "correct" answer, just adequate ones. And I wonder how we create quantitative, or at least comparative, expressions of network quality - how do we describe burst aspects (such as bursts of packet loss or the bunching together of packets?)  How do we separate foreseeable factors such as the solar blanking when a satellite transits the face of the sun as perceived from a particular ground station, from less predictable factors such as rain attenuating a satellite signal? My own sense it that we are going to need both of statistical expressions of quality and softer qualitative expressions. Perhaps we could perhaps measure user experience, eg can a movie be watched with less than N seconds of frame breakup?  (This can get far more "interesting" when we move from essentially one-way traffic, e.g. watching a video, to interactive/conversational, such as a voice call or multi-player gaming.) I won't ask, but I will allude to, the question of "who/what will enforce this?" OK, now that I've made my "hello" I will lean back and lurk and learn for a while...         --karl--