I can only assume that https://www.ietf.org/archive/id/draft-meow-mrrp-00.html exists to counterbalance the IPv8 thing.
grawity
hobbyist sysadmin and tech support person, still occasionally seen on IRC
interests: windows 98, macos 9/8, plan9, systemd, athena, x.25, all kinds of 80s-00s "retro" computing & networking
✨︎ no, I really don't have a good profile picture ✨︎
my favourite network is SUNET because in 2026 they still have good old L3 routing at every single hop and you can actually see the packet going through Sweden (e.g. to tungsten.libera.chat or umea.mirror.pkgbuild.com)
Came across this lovely collection of old-timey network mapping software screenshots while looking for WhatsUp alternatives
TP-Link's "Easy Smart" aka "Unmanaged Pro" series of cheap Ethernet switches (TL-SG108E, etc) technically has VLAN support, you can configure up to 32 VIDs as tagged/untagged.
But if you do so, the switch occasionally blasts its own DHCP discovers across *every single VLAN it knows about.*¹ Configured a port to have a tagged WiFi VLAN? Great, now the switch is going to pick up its management IP address from the guest subnet every couple of days.
I don't know who thought this would be a good feature.
¹ (Or maybe it chooses at random? Or maybe it's just not initializing the buffer and using whichever VID it saw in the most recent packet? Or maybe it's only doing that in the non-VLAN mode, the opposite of what I was assuming? Fuck knows.)
nullfs/failfs patches in vfs.git are interesting, apparently a process can now fchroot(FD_FAILFS_ROOT) and become filesystem-less, without needing a preexisting /var/run/empty or anything. like it doesn't even have a "/" anymore
a bit disappointed that @mikrotik@mikrotik.social developers now started replying to support tickets with wholesale LLM generated messages... but at least it seems the dev (and/or the LLM) finally understood the root of the issue, so there's that, I guess?
-
(thinking about it, I guess it's all the more jarring when the LLM output is still signed by the same dev's name as their own earlier messages. Like if I were to get a reply from "ClaudeGPT", then I would just set my expectations accordingly. But now the reply is from the same $PERSON_NAME as the 3 earlier messages, yet the text is *clearly* not written by them, and that's just bleh.)
I'm readingskimming through the final report on the last year's Spain grid collapse and huh that's a *lot* of graphs and math, and I didn't know they keep "oscilloscopic records" at like 400 Hz of Every Single Thing
https://www.entsoe.eu/publications/blackout/28-april-2025-iberian-blackout/#Publications_&_Documents
@domi@donotsta.re it's interesting how typing e.g. sqrt(warranty) is a syntax error but still gives you a license message
they should invent smartcards which just speak the ssh-agent protocol over USB
(I like my X.509 though, but it still seems like a cool idea)
reminds me of how the Bitcoin software in the early days used to find peers through joining a channel on [old] freenode...
"oh shit, the door is locked, how am I going to get my stuff now", I thought, while literally holding the door key in my hand and being the one who locked the door in the first place
Discord keeps dying mid-raid so we're trying out the built-in voice chat in WoW and it feels like we're dusting off a feature nobody has used in decades
I mean it works quite well (aside from the UI being five kinds of glitchy) but kinda has the vibe of going back to Win98 NetMeeting
@cks@mastodon.social for what it's worth, Feedly has not been able to receive any new posts since ~end of March, so now I have a cron job to update a local copy of cks.atom on my webserver and pointed Feedly to that instead
alright, time for rdo -S "sysctl kernel.unprivileged_userns_clone=0; rm -v $(modinfo -n rxrpc)"
this morning, on a day off, I get a syslog notification that my backup 4G modem isn't connecting. (it's at my little personal "lab" I have colocated at work.) I ssh to my syslog server to check what was happening, etc. It comes back up a minute later but I figured I'd scroll through logs anyway.
Literally *the next minute* after the 4G connection has come back up – while I'm still typing a grep command – the workplace's main fiber ISP goes down and 4G becomes my only remaining connection ;_;
I mean it's probably a coincidence, but.
"A keyserver MUST NOT support indexing or downloading certificates by prefix." [draft-gallagher-openpgp-hkp-09]
hmm what if a keyserver *did* support downloading PGP certificates by fingerprint prefix, but returned each item encrypted with its own fingerprint, so the client could only decrypt updates for the certs which it already has
my completely normal workflow:
-
trying to pick a good wallpaper for my WinXP laptop, as one does
-
browsing my homedir SMB share in XnView
-
'hmm, many of them still have XMP/IPTC/EXIF data', but XnView only shows EXIF, not XMP
-
double-click the .bat file on my desktop that rsh's to my home server and starts Xterm through the network
-
exiftool in the xterm shows a lot more details
-
:o
-
hit Ctrl-Shift-P to have SnagIt screenshot the xterm to the SMB share
-
Syncthing on home server picks up the screenshot in ~/Dropbox and syncs to the Microserver at work
-
too lazy to pacaur -y toot, so ssh to the server at work because I already have toot installed there
-
run chafa to display the screenshot via SIXEL to make sure it's the right one
-
toot post -m snag0012.png -d "alt text goes here" <<< "witty text goes here"
-
Ctrl-x ctrl-e the bash command to fix a typo before sending
I'm curious how communication happens between grid operators during a blackout. I guess cell phone towers (and the associated datacenters) keep running for a few hours? Not days?
(I've heard they run fiber along/inside of the transmission lines but not what exactly they use it for)
it seems that Neovim 0.12 "fixed" guicursor= so that trying to disable the feature as documented only enables it harder (always forcing the cursor to block shape)
as workaround I guess I no longer disable guicursor and let Nvim control it, but made the bash prompt reset it properly to "terminal default" afterwards (ESC [ 0 space q).
Also I was learning about the "DHCP Option 82" support in the less-cheap switches in order to track down a device that apparently wakes up every morning at 07:00, grabs a DHCP lease, then disappears for the rest of the day. By the time I'm awake it's already expired out of the MAC tables and everything.
so yesterday I went around enabling circuit-id support on various switch ports and I *think* now I know which of the (empty during summer) offices to check
salt-minion: "Unable to manage file: Jinja variable 'dict object' has no attribute 'id'; line 11"
me: salt \* cmd.run "apt purge salt-minion"
Feels kinda wrong to run an RDP client through X11-over-LAN, but hey, it works. Performance of FreeRDP being displayed over Wi-Fi to VcxSrv on WinXP is surprisingly good
(that is, much better than Firefox for example)