am i the only one for whom the slides are not loading?
", "time": "2026-07-23T12:12:18.000Z"}, {"author": "Muhammad Usama Sardar", "text": "me too. Slides are not loading in onsite tool.
", "time": "2026-07-23T12:12:58.000Z"}, {"author": "Dave Plonka", "text": "they are loading for me (onsite) using the full Meetecho client
", "time": "2026-07-23T12:15:18.000Z"}, {"author": "Stephen Farrell", "text": "when she's saying 'provider' does she mean web site?
", "time": "2026-07-23T12:16:43.000Z"}, {"author": "John Scudder", "text": "Did I correctly understand that EMD normalizes to \u201ceveryone is self-hosted\u201d?
", "time": "2026-07-23T12:17:14.000Z"}, {"author": "Ike Kunze", "text": "re slides not showing in onsite tool: probably due to the fact that it's a screen share, not a slide share.
", "time": "2026-07-23T12:17:52.000Z"}, {"author": "Jonathan Hoyland", "text": "Does having a long tail of thousands of providers that are each infinitesimally tiny make a difference to the EMD versus a small number of providers where the smallest one is equal to the same of the long tail?
", "time": "2026-07-23T12:17:55.000Z"}, {"author": "Yoav Nir", "text": "Stephen Farrell said:
\n\n\nwhen she's saying 'provider' does she mean web site?
\n
I thought she meant SDN or hosting service.
", "time": "2026-07-23T12:18:21.000Z"}, {"author": "Stephen Farrell", "text": "seems a bit of a mix of ASes/web properties
", "time": "2026-07-23T12:18:51.000Z"}, {"author": "Junxiao Shi", "text": "The term provider refers to a hosting provider.
", "time": "2026-07-23T12:19:22.000Z"}, {"author": "Stephen Farrell", "text": "thanks - I've not read the paper obviously:-)
", "time": "2026-07-23T12:19:41.000Z"}, {"author": "Paolo Saviano", "text": "\n\nre slides not showing in onsite tool: probably due to the fact that it's a screen share, not a slide share.
\n
correct
", "time": "2026-07-23T12:20:10.000Z"}, {"author": "Dirk Kutscher", "text": "yes
", "time": "2026-07-23T12:21:33.000Z"}, {"author": "Dirk Kutscher", "text": "Slides in the datatracker: https://datatracker.ietf.org/meeting/126/materials/slides-126-irtfopen-rumaisa-habib-formalizing-dependence-of-web-infrastructure-00
", "time": "2026-07-23T12:22:51.000Z"}, {"author": "Junxiao Shi", "text": "DDoS is the cause of Cloudflarization.
\nI used to have A+AAAA record directly to server, but a month of daily DDoS made me setup Zero Trust Tunnel.
Good point, @Junxiao Shi. This also allows vulnerable sites to open up to access to their content globally. (It is not unusual for an attacked site to limit access to their country otherwise.)
\n\"Cloudflarization\" is also interesting, to me, where it gets a formerly IPv4 site's content available on IPv6. I'm interested in determining how hypergiants are responsible for IPv6 adoption w.r.t. content availablity.
", "time": "2026-07-23T12:25:27.000Z"}, {"author": "Lixia Zhang", "text": "this was discussed in the DIN 2021 workshop report: \"... Again, as
\ntime went on, increasing network security threats, such as
\nviruses and DDoS attacks, together with shortage of IT manpower, resulted in web services being consolidated by the
\nbig cloud providers,...\"
Thanks, @lixia. It is still certainly true today, 5 years hence, if not more so. (IMO)
", "time": "2026-07-23T12:26:49.000Z"}, {"author": "Brian Trammell", "text": "this is the \"how big do you have to be to make it too expensive to knock you off the web\" question.
", "time": "2026-07-23T12:26:50.000Z"}, {"author": "Dave Plonka", "text": "I like it when globalization of content and IPv6 accessibility comes along for the ride, even when not the motivation.
", "time": "2026-07-23T12:27:20.000Z"}, {"author": "Dave Plonka", "text": "It allows intelligent evolution of hyperscalar infrastructure (like IPv6) to be easier to adopt.
", "time": "2026-07-23T12:27:51.000Z"}, {"author": "Maxence Fl\u00e9chier", "text": "the slides are slightly different though, but the gist of it is there
", "time": "2026-07-23T12:28:13.000Z"}, {"author": "Junxiao Shi", "text": "I don't feel guilty turning on the orange cloud.
\nMy origin server is located in a small provider (currently Advin Servers in Malaysia), so I can switch CDN providers if necessary.
right... good to remind that having your own origin is a key to being multi-CDN or nimble w.r.t. CDN.
", "time": "2026-07-23T12:28:55.000Z"}, {"author": "Junxiao Shi", "text": "I maintain No IPv6 Hall of Shame and IPv6 Less than /64 Hall of Incompetence of the hosting providers.
\nI don't think it's good to rely on CDN provider for IPv6 enablement, so I try to push hosting providers to have IPv6 to each server.
(eliot is asking one of the questions i deferred :) )
", "time": "2026-07-23T12:30:35.000Z"}, {"author": "Stephen Farrell", "text": "we could do with a 2nd independent LE for sure
", "time": "2026-07-23T12:31:09.000Z"}, {"author": "Stephen Farrell", "text": "but that's been true for years
", "time": "2026-07-23T12:31:28.000Z"}, {"author": "Jonathan Hoyland", "text": "Is there a meaningful way of combining the scores of the layers for a given region? What's the significance of being dependent on country A for layer X and dependent on country B for layer Y, vs being dependent on country A for both?
", "time": "2026-07-23T12:31:46.000Z"}, {"author": "Junxiao Shi", "text": "\n\n2nd independent LE
\n
There are other ACME providers available for free without manual registration.
\nCaddy server can acquire LetsEncrypt and ZeroSSL automatically.
Jonathan Hoyland said:
\n\n\nIs there a meaningful way of combining the scores of the layers for a given region? What's the significance of being dependent on country A for layer X and dependent on country B for layer Y, vs being dependent on country A for both?
\n
Or is there a hidden dependency on a single country or set of services? If country B relies on Country A....
", "time": "2026-07-23T12:34:35.000Z"}, {"author": "Jonathan Hoyland", "text": "@Eliot Lear So that's what I was thinking, you'd have to take the eigenvectors to figure out cascading dependencies
", "time": "2026-07-23T12:35:54.000Z"}, {"author": "Eliot Lear", "text": "But Rumaisa's right: it doesn't work out perfectly cleanly, because it depends on where the interface seams are.
", "time": "2026-07-23T12:36:35.000Z"}, {"author": "Jonathan Hoyland", "text": "Yeah, I think she's right too. You could run the computation easily enough, but what would the output mean?
", "time": "2026-07-23T12:37:19.000Z"}, {"author": "Gautam Akiwate", "text": "The _causal_ relationships would be the logical next step. For instance, we found that the choice of hosting provider often dictates the DNS provider and the CA.
", "time": "2026-07-23T12:37:23.000Z"}, {"author": "Gautam Akiwate", "text": "But that is really hard to reason about just looking at the scores. I am sure Rumaisa is game for follow up work if you folks have thoughts :)
", "time": "2026-07-23T12:38:18.000Z"}, {"author": "Jonathan Hoyland", "text": "That makes sense, Cloudflare has a default CA issuer for example, so the two are tied together.
", "time": "2026-07-23T12:38:19.000Z"}, {"author": "Eliot Lear", "text": "45 fragments?! What the heck is the MTU?
", "time": "2026-07-23T12:47:49.000Z"}, {"author": "Junxiao Shi", "text": "IPv4 allows fragments much smaller than MTU.
", "time": "2026-07-23T12:48:27.000Z"}, {"author": "Eliot Lear", "text": "Right but on a 1500 byte MTU 45 fragments is 33 bytes(!). So this was a very interesting use of fragmentation.
", "time": "2026-07-23T12:49:10.000Z"}, {"author": "Ryo Yanagida", "text": "maybe applying a really inefficient packet pacing :joy:
", "time": "2026-07-23T12:50:54.000Z"}, {"author": "Yoav Nir", "text": "Maybe trying to avoid DPI by pretending to be a SYN packet.
", "time": "2026-07-23T12:51:39.000Z"}, {"author": "Junxiao Shi", "text": "In-line DPI can choose to block all fragments, similar to some DDoS mitigation devices.
", "time": "2026-07-23T12:52:02.000Z"}, {"author": "Antoine Fressancourt", "text": "@Yoav then you end up being suspect of SYN Flood
", "time": "2026-07-23T12:52:35.000Z"}, {"author": "Yoav Nir", "text": "Antoine Fressancourt said:
\n\n\n@Yoav then you end up being suspect of SYN Flood
\n
I'm not making this up now. I've seen it tried. But that was 20 years ago. I don't think you can fool any modern DPI with this.
", "time": "2026-07-23T12:54:10.000Z"}, {"author": "Yoav Nir", "text": "But they are decrypting and re-encrypting the HTTPS traffic.
", "time": "2026-07-23T13:05:32.000Z"}, {"author": "Stephen Farrell", "text": "well for the .ru case they're also throttling
", "time": "2026-07-23T13:05:54.000Z"}, {"author": "Yoav Nir", "text": "Traffic is always throttled unless it's to https://www.speedtest.net
", "time": "2026-07-23T13:07:34.000Z"}]