← Feed

Posts

• Post #4377472

  The red halo of the red sun in the early morning. The ghosts of burnt forests, burnt houses. The new sight of summer in a world drying up.

• Post #4141791

  Morning visitor. I think this deer was born in our yard. Obviously he likes it here. We like him too, but we also like our flowers!

• Post #4018523

  This announcement of OpenAI chatbot escaping the sandbox an hacking an external database smells strange. It smells like some kind of demo rigged up for marketing purposes. Convince enterprises to pay a fortune for the AI service that is oh so powerful. On par with Anthropic advertising loudly that their new model should be reserved to a select few, and then proposing it to customers as an expensive option!

• Post #3767979

  Damn. Lindsay Graham was younger than me. Never would have guessed.

• Post #3664742

  The MLKEM discussion in the TLS WG has clearly reached popcorn status, at least for me. As in, take the flurry of partially informed drop ins and oft repeated arguments as a source of amusement. Relax, eat popcorn or drink wine or whatever, and enjoy the show. And appreciate that someone else has to disentangle all that...

• Post #2562294

  The "careful resume" RFC is finally published! I have been working on that issue for several years. The initial motivation came from the CNES in France and the ESA in Europe, who wanted a way to start QUIC connections faster on geo satellite links, instead of waiting multiple RTT for the congestion window to grow. And now we have it, all shiny and well polished by the IETF. https://www.rfc-editor.org/info/rfc9959

• Post #2465685

  The April Fool RFC on "protocol police" (https://www.rfc-editor.org/rfc/rfc9948.html) is in line with the classic IETF statement that "there is no protocol police". Funny, except that in fact there is. More and more big deployments use strict protocol checks, and a non compliant implementation does not get very far. Whether this is protocol police of "rule of the biggest deployments" is debatable, but the effect is similar.

• Post #1967223

  Nice PR landing in picoquic: push the number of supported connection from about 7,000 per core to over 100,000 by fixing the scheduler. Was a simple splay, sufficient in common scenarios, but it is O(logN) and that becomes an issue when the number of connections is larger than usual. Nice fix by Victor Stewart. https://github.com/private-octopus/picoquic/pull/2095

• Post #1878623

  @randulo @book Most words using "th" in French come from the Greek, e.g., théatre, thèse, sympathie, etc. The standard pronunciation for these words was set by Renaissance scholars who had no idea what ancient Greek actually sounded like, and assumed that Theta just sounded like T. French people learning English understand that the English TH is a foreign sound, not present in native French. They will try moving their tongue as the teacher explained, but it will often come out...

• Post #1044383

  @W6KME@mastodon.radio @ai6yr@m.ai6yr.org @darkuncle@infosec.exchange @nazokiyoubinbou@urusai.social You make me nostalgic. I should get a wheelbarrow like the one we had when I was a kid. All wood, except for the axle, and a strip of iron to circle the wooden wheel. Worked great. Ideal for carrying a load of manure...

• Post #912316

  @abhayakara@mastodon.nl The practical policing works by rigorous enforcement of MUST and SHOULD rules in RFC. This is sometimes very explicit, as in "Senders MUST NOT set the X bit to 1 if the Y bit is zero. Receivers MAY nuke the connection if they see this error." These clauses are vetted as part of the IETF process, and allow the protocol police to take harder actions than humming and frowning...

• Post #763569

  @mnot@techpolicy.social Chatting with a notebook about standards seems cool, and there are many reasons why geopriv was doomed, but what about the cheeses? I see camembert, sainte maure, some kind of époisse or similar, a Brillat-Savarin or its cousin, 3 kinds of gruyere, emmenthal or comté, a bleu that looks good, and then, a yellow cheese with a black crust? What is that? No idea! Please tell!

• Post #593342

  @carlmalamud@official.resource.org Bon voyage!

• Post #413779

  @Drbruced@aus.social That's exactly my experience with Microsoft's "copilot" when writing documentation. You start a sentence, and it "helpfully" propose a completion, which is its best guess of what I would like to write. Except it would also be the readers' best guess, and it would not bring them any new information. And thus the copilot guess would only be right if I was lazy and predictable and did not care for the reader.

• Post #274799

  @Ratoncita@mastodon.social @mister_shade02X2@mastodon.social @SeanCasten@mastodon.social There are two parallel things going on. First, there is the competition between solar, wind, fuel, natural gas and coal. Solar and wind and batteries are cheaper, and that's already impacting the worldwide revenues of oil and gas producers. The Trump administration is trying to preserve these profits by blocking solar and wind as much as it can, as in "ensure that energy production is profitable so...

• Post #266788

  @dave_andersen@hachyderm.io @SystemsAppr@discuss.systems Very true. I have seen "text book" examples of implementing Dijkstra's shortest path that were actually O(N^2)... And, no, I was not trying to dis math. That paper is definitely interesting, even if they did not do the best job about modestly acknowledging limitations. Which student does that?

• Post #245920

  @woody@pleroma.pch.net Makes one think of Lamborghini...

• Post #189249

  When we look at the performance of QUIC, we often find that the limiting factor is CPU consumption of socket calls and crypto processing. It is, but congestion control also matters quite a big. In this blog (https://www.privateoctopus.com/2026/01/30/cpu_bound.html) I explain how fixing the "max RTT" measurement and the pacing algorithm for the C4 algorithm improved tests of picoquic on loopback from "worse than BBR" to "better than Cubic".

• Post #167257

  @woody@pleroma.pch.net "Laser" is one way to say it -- no light involved We are looking here at phased arrays of small speakers, all sending the same sound wave, each applying a delay so that the sound waves superpose and add at one exact point in space. I remember a demo of that tech at a Microsoft Research fair years ago. The effect is weird. The target hears the sound, nearby people just hear low level noise. By the way, the same principle also works for sound capture.