| 11:12 | asciilifeform | logs.bitdash.io loox down |
| 11:12 | billymg | bitdash.io is back up and running, along with the trb node on that box. logger should be working too (let's see if this message comes through). trilema.com still down, wonder who's in charge of that box these days |
| 11:12 | asciilifeform | ( billymg still tuned in ? ) |
| 11:12 | billymg | asciilifeform: got hit by a russian missile https://status.itldc.com/incident/188 |
| 11:12 | billymg | (or drone) |
| 11:12 | signpost | france24.com/en/europe/20260925-trump-made-final-decision-to-give-ukraine-patriot-licence-zelensky-says << the meat grinder hungers. |
| 11:12 | billymg | looks like trilema's box was hit by the same russian missile -- trilema.com has been down since then |
| 11:12 | asciilifeform | potentially it could be so that the host doesn't even need own pestron. simply sees connection to what loox like regular e.g. ssh. |
| 11:12 | asciilifeform | for bonus lulz, afaik (in so far as the open/leaked lit suggest) this is roughly how usg's own 'siprnet' wankage worx. (with 1980s nsa atrocities, naturally) |
| 11:12 | signpost | an item that is more/less wireguard with roaming and NAT punching would be nice. |
| 11:12 | signpost | would you do the wrapped networking as a red packet type, or do you think there's more to it? |
| 11:12 | asciilifeform | http://logs.nosuchlabs.com/log/pest/2026-10-27#1036266 << depends whether the idea is simply a filter in front of a pestron on the host, or, instead, 'iron pestron', neh |
| 11:12 | asciilifeform | meanwhile, re http://logs.nosuchlabs.com/log/pest/2026-10-23#1036251 >> days, not yrs, lol : https://arstechnica.com/security/2026/10/code-execution-flaw-in-uefi-firmware-threatens-most-servers-on-planet-earth/ |
| 11:12 | asciilifeform | ^ admittedly, 'weak sauce' magickpacket -- has to arrive during boot, and on box w/ pxeism switched on. but apparently there's over9000 of'em. |
| 11:12 | asciilifeform | ... where erryone who is reading, is also thereby 'hosting' |
| 11:12 | asciilifeform | otherwise it's a 'last 1 plz turn off the lights' situation. |
| 11:12 | signpost | 100% agree that human civ requires there to be a navigable public space |
| 11:12 | signpost | testing send through the butterstick |
| 11:12 | asciilifeform | signpost: seems to work. what's it do to outgoing packets tho? ( from earlier thrd, seems like it was intended as a filter for incomings ? ) |
| 11:12 | * | asciilifeform wonders how wot/at is config'd on the fpga stick |
| 11:12 | * | asciilifeform predicts that the legendary 'magic nsa packet' 'bug' will be found in not too many yrs, and the pill, at least in so far as private boxen are concerned, is logically this and only this, and anyone not already using pestronic ssh will have only himself to blame |
| 11:12 | asciilifeform | ( see e.g. & elsewhere ) |
| 11:12 | signpost | yeah, that's why I'm hitting the gas on this. I figure chicom oss models diddling all our realteks, intels, etc. on a 3090 are imminent. |
| 11:12 | signpost | butterstick is handling 2-way encryption / decryption and hashing |
| 11:12 | signpost | a computer sits behind with the butterstick acting as its nic. there's a small protocol to interact with the butterstick over usb |
| 11:12 | signpost | https://problematic.site/src/v/tempest/patches/tempest_box_usb_backend_signpost.vpatch << beware of claudism |
| 11:12 | signpost | intentionally unsigned for now |
| 11:12 | signpost | will get the fpga stuff uploaded in a bit |
| 11:12 | asciilifeform | oh hm i was assuming the thing enumerated as a regular usb nic (simply one that only hands off valid red'd pest packets) rather than 'driven stick' by pestron from userland |
| 11:12 | asciilifeform | ^ in that variant the host machine not even needs to know the thing aint a standard usb nic |
| 11:12 | asciilifeform | thinking further along this line, in the above variant, the wot could even be controlled via another set of pins (serial), plugged into yet another machine, or untouched most of the time. and literally all the host sees is a regular usb nic... |
| 11:12 | signpost | defo could be the right finished item. for now, got crypto and hashing right on the fpga. it has been up since the other day without incident handling black packets. |
| 11:13 | signpost | I can see a reason for putting the encoding on the outside of course, but wasn't going to invert the protocol myself. |
| 11:13 | asciilifeform | 3rd party can't poison'em |
| 11:13 | signpost | you at least get the uncoordinated swarm download out of it, which was at this stage all I was aiming for |
| 11:13 | asciilifeform | lubying is for 'grown-up www', i.e. public hosting invulnerable to ddos & diddle |
| 11:13 | signpost | can ask P peers for N encoded packets of $hash payload without them coordinating with each other |
| 11:13 | asciilifeform | right. talking to N peers costs times-N tho. |
| 11:13 | asciilifeform | still potentially a win vs. asking that-one-guy for 10GB and hoping he stays online tho |
| 11:13 | signpost | current model is the requester apportions his packet requests politely to each peer, something like 103% of payload size / peer count, per peer |
| 11:13 | asciilifeform | what asciilifeform was shooting for was the thing where 'there is no www host', i.e. the 'url' is a pubkey+hash and whoever got it, can send you chunks, any of which individually are immediately and correctly sorted into valid vs crapola, and once you have $magicnum of'em, guaranteed correct reassembly. |
| 11:13 | signpost | yup, same problem isn't it? what I did was introduce a protocol verb in pest for asking peers for a hash, then asking for streams of that hash. |
| 11:13 | asciilifeform | exactly same, other than in the 'public' (vs pestronic) variant, 3rd party can throw arbitrary constructed crapola at the receiver, and it gotta reject it with 'nothing to the stranger', i.e. w/out taking any longer than 1-packet-at-line-rt, or 'extending credit' for any memory beyond the 1-packet buffer in the nic |
| 11:13 | signpost | mhm, got it. |
| 11:13 | * | asciilifeform was hoping to discover a pill where the lubytron *is* the authenticator. but seems as if this necessarily requires hash trees (a la merkle) and introduces carouseling. so nope. |
| 11:13 | asciilifeform | (whereas if there is any possibility of the lubytron being fed shit, it will necessarily fail reconstruct, even from 1 packet o'shit) |
| 11:13 | asciilifeform | (what's wrong with carouseling? well, if enemy can make it so that e.g. 99 of 100 packets are garbage, the receiver may wait for the carousel potentially 4evah) |
| 11:13 | asciilifeform | the packetz gotta be idempotent. |
| 11:13 | signpost | yup, if there is a scheme whereby one can avoid having the all-or-nothing encrypted outer wrapper per packet, I am too dumb to see it. |
| 11:13 | asciilifeform | tldr, ther is, but makes tx $bitness-times moar expensive. |
| 11:13 | asciilifeform | tldr-x2: afaik nobody, of who's tuned in, orig met in meatspace. there has to be a hygienic way to post publicly -- or yer in 'bunker'. |
| 11:13 | asciilifeform | bunker w/ whoever you knew from decades ago, minus the burned-out folx. which lim -->0 over t |
| 11:13 | asciilifeform | this means, concretely, that one oughta be able to go from e.g. pgp key (that ideally has a vtronic history or otherwise anchored in reality) to the equiv of a public www |
| 11:13 | signpost | looks very much like an intentional op against having a public internet |
| 11:13 | signpost | I had to be cooking on multiple cores on a ryzen 7 to approach gigabit streams with occ. |
| 11:13 | asciilifeform | occ ? |
| 11:13 | signpost | online-codes c, my lib that implements the maymounkov algo |
| 11:13 | asciilifeform | aa |
| 11:13 | signpost | encoders can all piss independently; decoders have to coordinate to not trip over one another as they peel the payload together. |
| 11:13 | asciilifeform | (naturally 'public' broadcast packets, if there's actual link security a la pest, ~0 of this is needed) |
| 11:13 | asciilifeform | if wanking over e.g. radio, the packets sigs double as integrity checks, obv. |
| 11:13 | asciilifeform | (just as pest seals do) |
| 11:13 | signpost | yeah, I just rode in pest black packets, created a separate red packet type for them. |
| 11:13 | asciilifeform | the thing is, there aint imho a huge amt of point to lubying transmissions that are already blacked via pest |
| 11:13 | signpost | those tuned in should check their php-serving boxes in light of today's RCEs dating back to 2004 |
| 11:13 | signpost | in related news, I am working on making my pest node (tempest) available as vpatches. it is *not* hand-spun tmsr-ist C; it's claudism. with that said, I have separated the thing into a black node and red node, and the black node has two implementations, software, and something that runs on a butterstick. |
| 11:13 | signpost | the thing is a USB NIC that lets only properly decrypted red packets arrive at the machine behind it. |
| 11:13 | signpost | seems like such a thing needs to exist. |
| 11:13 | signpost | (and of course, lets only properly encrypted black packets leave) |
| 11:13 | asciilifeform | re: php clusterfuck: asciilifeform for 20y nao searched for a usable replacement for that pile o'shit, and still all earz for practical ideas. |
| 11:13 | asciilifeform | (inb4 'static html' -- plz no) |
| 11:13 | asciilifeform | phpism is imho pretty much morally dead, even before all the installs get 0d'd and start hosting cock elongator sp4m this week. asciilifeform turned off comments a while back, the llmtrons solved the comment rebus 2+y ago |
| 11:13 | asciilifeform | if anyone has ideas re: a pill for this, that don't reduce to 'whack-a-mole' -- plz speak up. |
| 11:13 | signpost | nfi, hence focus on getting at least some of my machines behind a pestronic membrane |
| 11:13 | signpost | a friend's grandfather's line: "nature extends a lot of credit but never forgets to send the bill". seems like it's here for open-sores. |
| 11:13 | asciilifeform | buttoning up ssh etc is well an'good, but oughta be possib to host a public www, shouldn't require a cyclophasotron lol |
| 11:13 | signpost | should! and edible food at the grocery store also |
| 11:13 | * | asciilifeform plugs in spiritism board & summons mp, 'if enemy wanted to get people to wall themselves up in bunkers with only sisters to fuck, they're succeeding' |
| 11:13 | asciilifeform | ... best it was able to cough up was rabin-williams. cheap verify, $bitness-times moar expensive per packet to sign |
| 11:13 | asciilifeform | i.e. needs ~8fold bag o'packets, pre-genned, coupla ~hrs per GB, approx |
| 11:13 | asciilifeform | the 'verification turd' (i.e. equiv. of bittorrent's magnet link) for subj would be, minimally: 1) the rw pubkey (use 1ce) 2) hash of final payload. |
| 11:13 | asciilifeform | under this, luby, or turbo, or whatever rateless coad. |
| 11:13 | asciilifeform | ^ not earth-shaking discovery. simulated the bare bones scheme w/ the bot, demonstrated (no prizes for guessing) e.g. ec crypto very certainly too slow to keep up at any reasonable line rate (just as asciilifeform said when folx were chewing on what turned into pest) |
| 11:13 | asciilifeform | e.g. typical 1core arm can, just barely, verify 2048bit rw at line rt, if sewed in asm |
| 11:13 | asciilifeform | (fughet about encoding anywhere near line rt tho) |
| 11:13 | * | asciilifeform was stabbing at the problem in the general case, i.e. where all packets (w/ exception of the pubkey + finalhash) are idempotent, i.e. there is no 'carousel wait' as there would be with a merkle tree scheme |
| 11:13 | asciilifeform | the notion being that you can give someone a small turd that can then be used to piece a payload outta certain number of packets, recv'd in any order, and there's 0 possib of bt-style poisoning |
| 11:13 | asciilifeform | signpost: will defo read |
| 11:13 | asciilifeform | arguably last 20y was this 'op', lel |
| 11:13 | asciilifeform | anyways the pill here is signpost's luby widget. /me recently fed the problem of 'possible to authenticate broadcast luby packets one at a time?' to industrial llmtron, and... |
| 11:22 | billymg | the host's latest update on the matter https://vsys.host/news/recovery-update-servers-and-data-returning-new-locations-coming |