Aaron A. Glenn @aag@bsd.network

I will be tooting into this void as a form of "rubber duck" debugging and catharsis.

you wont hurt my feelings by unfollowing; but I will try to be somewhat entertaining in expressing my trials and tribulations -- and of course appreciate any cluebats you can provide!

Hi. I'm a long time (ethernet/IP) packet networking nerd. I am easily excited by functional programming (OCaml, F#), unikernels (any kind!), ISAs that aren't x86 or ARM (PowerPC & MIPS), and domain specific languages for domains I'm interested in (think packet forwarding & control planes -- P4, propane, cocoon, NetKAT).

I'm not doing hash tags so hopefully keywords are enough for being discoverable on this federated toot streaming service.

what I really *need* to work on, after completing things started, is looking at how to integrated P4Runtime code into my very, very hobbled attempt at integrating a P4 pipeline into Cilium with p4c-xdp.

"cool, I can replace firewall rules and write network policies with code (P4)" is about 20% as cool as it could be, without the control plane (P4Runtime) part.

and that has a very hard deadline of August 10th.

<<this concludes my self-hating rant>>

and that's not even half my list.

but instead of *completing* of those things I've started (with great momentum and interest, mind you) I'm looking at *another* thing on my "TODO" list.

"hmm, let's get SDKLT & XGSSIM running somewhere and poke around that!"

NO AARON
NO

YOU HAVE TO FINISH THINGS

but I dont waaaaannnnnaaaaaaaaaa

I got Juniper vMX VMs I am trying to spin up in QEMU on OpenBSD -- to gain some experience with Advanced Forwarding Interface, streaming telemetry, & for things like BGP & BFD "tests"

I got gprof(1) OpenBSD & NetBSD kernels *built*, ready to start grabbing codepath traces for the network stack.

I drank beer with @phessler *weeks* ago & got answers to EUI64 address generation (why does it skip vio interfaces?!)

I have *four* P4 blog posts 60-80% done.

this is maddening.

is there a medical diagnoses for someone who just can't, for the life of them, *complete* anything? I finish stuff -- answer the original question/thought that led me to work on something -- but have literally now seven things I haven't *completed*

I'm beginning to think, now at the age of 34, that something isn't all right in my head.

do I keep forgetting the great rush of accomplishment feels like when I've completed something? how do I keep that feeling front and center??

The state being Massachusetts.

https://twitter.com/matt_cam/status/1010175113583628288?s=17

I'm on a train in sunny Berlin reading ICE agents are doing mass raids at court houses after the state Supreme Court ruled detainment papers are warrants (because they aren't...never have been) and Pompeo is signalling the US is going to give up on Ukraine sovereignty in exchange for who-knows-what.

The world order is crumbling. Fast. Everything is going to look dangerously different by 2019.

I didn't even get to do 30% of what I wanted (and frankly, they needed) to do on the last "real" network I worked on.

it's been six months, and I have no dearth of interesting, network related things to occupy myself with...but, I somehow miss "it"

working on and solving problems, end to end, on a *physical* network.

part of me wants to start cold mailing other networks that used this DOCSIS over fixed wireless technology.

I'm just saying, Multicore OCaml + MirageOS (unikernel, <10ms boot time) & Jitsu (TCP proxy, unikernel launcher) + ARM64 and suddenly VNFs at the "edge" look downright impossible not to deploy.

Gee, I wonder what the best language to write VNFs in might be...🤔

You don't appreciate how broken the Python packaging system is until you work with the nixpkg tools trying to assist you with it.

I say this wearing a RUN BSD shirt:

NixOS and Alpine have been the most pleasant experiences installing Linux I've ever had.

Tune in later when I attempt to actually use either of them to do something and begin ranting and raving as is tradition.

both Cisco and Arista are selling Tofino based switches.

I don't feel so bad impulsively buying p4.solutions now

https://www.pcworld.com/article/3187576/servers/its-not-the-end-of-sparc-chips-yet.html

so who wants to crowdsource one of these bad boys? I'll find some colocation for it.

why is pandoc practically impossible to install?

I thought Haskell was elegant? attempting to install pandoc via cabal is a mess and a half.

I have been grabbing route-views and PCH MRT files, converting them to BMP (why? because why not) and dumping them into Clickhouse.

so of course I'm back on my bullshit and registered dfz.watch

because why not -- domains are internet real estate!!!1

If hyper threading and all these other glaring security holes, esp for virtualized environments, were known back when SPARC and other ISAs weren't dead yet...do you think they'd have still died out?

Seems like all the x86 speed gains touted as a reason for Intel superiority over competing ISAs turns out to be complete BS

Can we get SPARC back?