I actually already submitted a while ago. There’s a lot of additional work I want to do though, like better docs (especially about how config work), adding command history and improving caching and my VT100 support.
VR is gigajank at the moment. Everything is still very much in the infancy stage. However when you get something that really works, it feels really amazing. The big thing I’m looking forward to is the rumored Valve Index 2 which will have some kind of standalone mode.
I am on vacation this week, so my activities might vary daily as the mood strikes! My wife did buy me a couple of courses on Ansible and Pycharm usage for Christmas so I’ll probably dive into those. I also need to take a look at my (neglected) personal infrastructure and run in updates and other maintenance.
End of the week sees us celebrating New Year’s Eve with our best friends, that back to work on Monday :/
Back home after a lovely week of vacation, and feeling pretty great.
Work: Figuring out how to get data in and out of weird semi-proprietary map sharing systems. Looks like most of it is kindasorta broadcast/multicast UDP carrying protobuf payloads, and unlike the last time we looked at it there’s actually some docs now. Could be actually fun, or could be a giant pain, we will see! Never mind, looks like I’m keeping a terrible old app staggering along for another month so that it can continue to be terrible instead of getting ahead on work so I can try to replace it with something less terrible.
Nonwork: Starting to port my proof-of-concept type checker into my compiler and seeing if it actually works, which it should. Also dealing with various accounting things, writing new year’s letters, and other general year-end business.
After six years of working for GitLab, this week is my last week. Starting 2022, I’ll be working on Inko full-time (some extra info). What I’ll be doing this week is mostly offboarding, and slowly ramping up work on Inko.
I got the last star. My work-in-progress from the 19th was only a few lines of code away from working. Then it was just optimizing to get it from 20+ minutes down to 4 seconds before I felt like I was really done.
Finally regained enough energy to actually do something instead of just reacting to things happening everywhere (I need to reconsider how I work). So this week, I’ll be mostly learning Common Lisp!
3 weeks ago, I was sure I could write a blog post about my favorite Shell (Bash) library in an evening (now that I’ve finally written something short+real+released+digestible that uses it).
Hopefully I’ll finish the multipart series that has scope-crept into… :)
Spending my vacation days configuring my home network with a Wireguard VPN. I had setup my home network with just simple port-forwarding, which has not been ideal in terms of security, having my public IP exposed with a handful of open ports. Now that’s about to change.
Hopefully I also get a chance to continue my Ansible playbooks as well. Really blown away how great Ansible is!
Learning the BEAM and OTP by porting code from Learn You Some Erlang to Elixir. Motivated partly by the OTP feeling like a bit of a white whale in terms of actually understanding it enough to be able to reason about it, and partly by wanting port my polling FS watcher from Powershell/Janet to Elixir, for use in keeping me from feeling like building websites outside of Elixir is too clunky.
I know that existing FS watchers exist, but I’ve not felt comfortable with OTPs start_link/init combo. For lightweight processes, it has felt like a semi-decent amount of ceremony behind it all. So, by porting this exercise to Elixir, I’m hoping to get more comfortable with links, monitors, receives, spawning, and so on.
Lots of planning and review: annual, quarter, biweekly. Just like Year Compass, my BuJo, the like. Suggestions welcome, especially when it comes to picking out the choice systems amid the countless offerings out there.
Configuring a virtual bridge for hundreds of netns separated veth interface pairs for our project simulator system.
Have a stupefying case when they all send UDP broadcasts (approx simultaneously), the packets on the bridge get source address of the first veth to come up but the correct payload. This is why am not a sysadmin!
Fighting against Docker on my laptop: when I start container with bridged network, Docker install a default route on the veth and cut me off from Internet. It seems related to internal Docker’s DHCP server.
Keep playing with Serverless framework
Trying to understand how a GAN is working and training my first one.
My goals remain the same as the weekend note, but I wanted to update: I’m feeling good, and wrote some bad-but-functional golang code that I’m iterating on!!!: https://github.com/neeasade/gott
I am proceeding with half-time job grind while taking the spare time to investigate procedural content generation methods in video games. This time about constraint satisfaction problems and their usage in building interiors, along other search-based and grammar-based algorithms and their usage.
Hopefully have some spare time to play around my newly received Stem Player toy (made by Kano) and work on my RGB peripherals and devices synchronization desktop app, RGBMaster.
Really curious how you get on with ssdnodes. I don’t see how they can offer the prices that they do . Some comments elsewhere suggest that they do it by massively oversubscribing everything, including RAM (most cloud providers give you dedicated RAM, it looks as if they do host-based swapping, which is fine for bursty workloads but will suck if you’re actually using all of the RAM).
If they aren’t oversubscribing, they’d need you to rent the same hardware from them for over 10 years to break even, assuming that their electricity and storage space is free. 48 GiB of RAM costs on the order of $400 (assuming that they aren’t using ECC, which they really should for shared hosting or it’s really easy for a tenant to escape via RowHammer). 720GiB of NVMe is around $50-100. 12 threads on an E5 Xeon are adds up to around $700. Add in a motherboard, case, and so on, and you’re looking at a minimum of around $1000 - $1500 to be able to buy the hardware to support their KVM / 2X-LARGE offering without oversubscription. If they make back $100/year then they’re going to need to oversubscribe RAM and CPU by a factor of around five to be able to recoup their initial hardware investment in three years. If they’re covering their running costs and making a profit then they’re likely to be oversubscribing by a factor of 10 or more.
It’s possible that they’re burning through VC money at a high rate to try to grow market share but if they aren’t then they’re definitely in the ‘too good to be true’ category.
They basically say they oversubscribe in their marketing materials, but that’s fine for this use case. If I had enough users for that to matter, I wouldn’t be penny pinching.
Bought two Chromebooks and have been playing around getting Linux installed but went back to ChromeOS when I realized if I wanted a true Linux machine I need to flash the firmware. So back to the container within the VM. Waiting on some bits to put together a makeshift SuzyQable since the supply chain crisis has made the orange ones hard to come by. Once I have that working (with just a tiny bit of soldering, thank God) I’ll hopefully be able to flash the firmware on one of them so as to have that aforementioned machine. After that, who knows! Maybe give it away to someone or use it as a beater. Otherwise, having not used a Chromebook in years this low-end HP model has been surprisingly pleasant to setup and use and has pretty sweet speakers for those Steve Roach ambient classics of the 1980s.
For work, I’m writing up what our dev team did over the last year to give perspective hires some info on what our worklife is like. I work for Norway’s biggest subscription newspaper, and our big highlights: redesign of the frontpage to provide users with more context and increase time well spent when browsing us. Work with an ML trained version of our podcast journalists which will provide text-to-speech for all articles. Using AI to generate real estate trend articles. And a stream dedicated to working with innovation in a lean manner. We also spent one day every fortnight working on self-improvement, ranging from ML, Git, React, language development. If that sounds compelling to you, we’re hiring people of all levels in the job add here. Feel free to mention Lobsters in your cover letter.
In my private time, I’m working away on Derw. I have a monthly blog post coming up going over all the major changes, along with some thoughts and reflections on language design. I’m getting very close to the point where I can start my first big project in Derw itself: a homepage.
I have the week off from work, so I’m planning to spend a full 40-hour week just working on my social media protocol. Currently I’m focused on a markup language optimized for social media; some people would be fine with plaintext, but there needs to be a markup language otherwise someone’s going to shove HTML in there, and then it’s all over.
Today I revisited my goals and I’m pretty happy with how it all factored out: Core goals, goal implications, strategies, and then the beginnings of the spec.
(The larger project, Cavern, is a social media protocol with privacy and user empowerment as main goals. It uses an interesting store-and-forward design that allows use of fairly low-tech hosting. It’s my bid for quasi-self-hosted social media that’s nevertheless user-friendly.)
I’m migrating a Rust web service from Rocket to actix-web. It seems to be going well but like most big rust migrations it’s been a day of compiler errors. I’ll probably slow down after vacation but hopefully I can finally get it to compile again and ditch Rocket.
Then write a blog post about my pros and cons of Rocket.
Yes. Although to be honest I’m not even using much of the actix framework. I’m doing a bunch of stuff myself.
I guess read the blog post for the details… but my main problem with Rocket is that forcing FromRequest usage doesn’t work for me. It is cute and convenient for small examples but it is too limited for a lot of stuff. For example I wanted to add ETag support to a bunch of pages but you can’t do this from a FromRequest implemention (you can only return a status code, not modify the response or generate a custom response) so I would need to do it in the method body, but without being able to have a Request object each of these “middlewares” needs to have its own set of arguments pulled from the request which is a very awkward API. I find actix-web with user-accessible Request and Response objects to be far more compostable. Maybe I’m missing how you are supposed to implement something like this but I think it comes down to using regular code is always going to give you more flexibility than some small FromRequest API where the only per-handler config is at the type level.
I will miss the URL generation from Rocket though. It was really nice to have type-safe link generation but I’m manually making a rough solution for now which is good enough, maybe I’ll try to push something upstream to actix at some point.
Deeply enjoying the book “Word by Word” by Kory Stamper. Best-written (at the word level) book I’ve ever read; unsurprising, because the author is a dictionary writer (sorry, lexicographer). Also, getting ready to perhaps hear a lot about antisemitism in Poland for some stuff I’m doing. Maybe it’ll expand my horizons.
I’m finally getting around to finishing the painting of the walls in my house. Other than that, I intend to get started with stronglifts 5x5 and generally recharge and relax so I can get back to work refreshed in the new year. And perhaps I’ll take some time to work on CHICKEN as well.
Working on blog posts listing best SFF books I read this year, reviewing a technical book I’ve been reading this month, may be a post reviewing my own work this year, etc.
So as it is said in the .plan - finishing all tasks as last week I learned that my current contract will not be extended, so I will be looking for new contract in new year. After that I need to refresh my CV and start looking for new workplace.
I’d like to teach myself some basic GC algorithms: refcounting, full collection, generational collection, etc.
Anyone have any good resources to self-teach these? I can look up discussions of them, I suppose I’m looking for a bit more exposition than commented code.
Nothing at all.
Edit: Unless we get unpleasant email to security@. Wish me luck.
I say this with feeling, good luck!
I am finishing up my Ada crate of the year entry.
Goodluck! Post your entry when you’re done :)
I actually already submitted a while ago. There’s a lot of additional work I want to do though, like better docs (especially about how config work), adding command history and improving caching and my VT100 support.
More VR rhythm game streaming and probably writing up what I’m excited for in 2022!
Yeah a write up would be nice. I have been holding out on VR for years as I am not sure if it worth the value just yet
VR is gigajank at the moment. Everything is still very much in the infancy stage. However when you get something that really works, it feels really amazing. The big thing I’m looking forward to is the rumored Valve Index 2 which will have some kind of standalone mode.
I am on vacation this week, so my activities might vary daily as the mood strikes! My wife did buy me a couple of courses on Ansible and Pycharm usage for Christmas so I’ll probably dive into those. I also need to take a look at my (neglected) personal infrastructure and run in updates and other maintenance.
End of the week sees us celebrating New Year’s Eve with our best friends, that back to work on Monday :/
Back home after a lovely week of vacation, and feeling pretty great.
Work:
Figuring out how to get data in and out of weird semi-proprietary map sharing systems. Looks like most of it is kindasorta broadcast/multicast UDP carrying protobuf payloads, and unlike the last time we looked at it there’s actually some docs now. Could be actually fun, or could be a giant pain, we will see!Never mind, looks like I’m keeping a terrible old app staggering along for another month so that it can continue to be terrible instead of getting ahead on work so I can try to replace it with something less terrible.Nonwork: Starting to port my proof-of-concept type checker into my compiler and seeing if it actually works, which it should. Also dealing with various accounting things, writing new year’s letters, and other general year-end business.
“Deathloop”
I’ll probably also go to the Vizcaya museum and take pictures of lizards at least once too.
After six years of working for GitLab, this week is my last week. Starting 2022, I’ll be working on Inko full-time (some extra info). What I’ll be doing this week is mostly offboarding, and slowly ramping up work on Inko.
I want to make a comparison of how you build a web application with different frameworks.
The idea is to have a self explaining shellscript for each framework which sets up a web application with routing, templates and users accounts.
I am halfway done with the one for Django:
https://www.gibney.org/from_debian_to_web_app
I still have 3 stars to get in Advent of Code. Friggin Day 19. No spoilers, please.
I’ve got 18 stars to get. I want them though, and don’t think I’ll keep on it after December. I didn’t in years past.
I feel you. Working on the same problem here.
I got the last star. My work-in-progress from the 19th was only a few lines of code away from working. Then it was just optimizing to get it from 20+ minutes down to 4 seconds before I felt like I was really done.
Finally regained enough energy to actually do something instead of just reacting to things happening everywhere (I need to reconsider how I work). So this week, I’ll be mostly learning Common Lisp!
Working on yet another NixOS deployment tool and maybe unifying my system configurations.
Long-time lurker here. I’m creating a visualization of all 11K sound-effects in Super Smash Bros Ultimate
3 weeks ago, I was sure I could write a blog post about my favorite Shell (Bash) library in an evening (now that I’ve finally written something short+real+released+digestible that uses it).
Hopefully I’ll finish the multipart series that has scope-crept into… :)
Doing some dangerous refactors while the codebase is quiet
Spending my vacation days configuring my home network with a Wireguard VPN. I had setup my home network with just simple port-forwarding, which has not been ideal in terms of security, having my public IP exposed with a handful of open ports. Now that’s about to change.
Hopefully I also get a chance to continue my Ansible playbooks as well. Really blown away how great Ansible is!
Visiting friends, going away with the other half for our tenth anniversary and not doing a lot is my current plan.
Happy anniversary!
Learning the BEAM and OTP by porting code from Learn You Some Erlang to Elixir. Motivated partly by the OTP feeling like a bit of a white whale in terms of actually understanding it enough to be able to reason about it, and partly by wanting port my polling FS watcher from Powershell/Janet to Elixir, for use in keeping me from feeling like building websites outside of Elixir is too clunky.
I know that existing FS watchers exist, but I’ve not felt comfortable with OTPs start_link/init combo. For lightweight processes, it has felt like a semi-decent amount of ceremony behind it all. So, by porting this exercise to Elixir, I’m hoping to get more comfortable with links, monitors, receives, spawning, and so on.
Lots of planning and review: annual, quarter, biweekly. Just like Year Compass, my BuJo, the like. Suggestions welcome, especially when it comes to picking out the choice systems amid the countless offerings out there.
Configuring a virtual bridge for hundreds of netns separated veth interface pairs for our project simulator system.
Have a stupefying case when they all send UDP broadcasts (approx simultaneously), the packets on the bridge get source address of the first veth to come up but the correct payload. This is why am not a sysadmin!
Fighting against Docker on my laptop: when I start container with bridged network, Docker install a default route on the veth and cut me off from Internet. It seems related to internal Docker’s DHCP server.
Keep playing with Serverless framework
Trying to understand how a GAN is working and training my first one.
Post the gan resources you’re using?
I am using this one. It’s an already prepared GAN, I am not so skilled to write my own GAN.
The tutorials on Tensorflow and Keras documenation or this
My goals remain the same as the weekend note, but I wanted to update: I’m feeling good, and wrote some bad-but-functional golang code that I’m iterating on!!!: https://github.com/neeasade/gott
I am proceeding with half-time job grind while taking the spare time to investigate procedural content generation methods in video games. This time about constraint satisfaction problems and their usage in building interiors, along other search-based and grammar-based algorithms and their usage. Hopefully have some spare time to play around my newly received Stem Player toy (made by Kano) and work on my RGB peripherals and devices synchronization desktop app, RGBMaster.
Completed Andrew Ng’s Deep Learning course 1 on Coursera, today. Probably start on the next course this week.
Also hoping to finish up the next post for https://dawn-lang.org, formalizing the Untyped Multistack Concatentive Calculus. The corresponding toy language is pretty much done: https://github.com/dawn-lang/umcc/
Currently, working on migrating https://sneakysnake.io from digital ocean to ssdnodes. $300 for 3 years of a 12vCPU server was too good to pass up.
Other than that, going to a Toadies concert tomorrow night, and NYE boat party Friday night 🎉😃
Really curious how you get on with ssdnodes. I don’t see how they can offer the prices that they do . Some comments elsewhere suggest that they do it by massively oversubscribing everything, including RAM (most cloud providers give you dedicated RAM, it looks as if they do host-based swapping, which is fine for bursty workloads but will suck if you’re actually using all of the RAM).
If they aren’t oversubscribing, they’d need you to rent the same hardware from them for over 10 years to break even, assuming that their electricity and storage space is free. 48 GiB of RAM costs on the order of $400 (assuming that they aren’t using ECC, which they really should for shared hosting or it’s really easy for a tenant to escape via RowHammer). 720GiB of NVMe is around $50-100. 12 threads on an E5 Xeon are adds up to around $700. Add in a motherboard, case, and so on, and you’re looking at a minimum of around $1000 - $1500 to be able to buy the hardware to support their KVM / 2X-LARGE offering without oversubscription. If they make back $100/year then they’re going to need to oversubscribe RAM and CPU by a factor of around five to be able to recoup their initial hardware investment in three years. If they’re covering their running costs and making a profit then they’re likely to be oversubscribing by a factor of 10 or more.
It’s possible that they’re burning through VC money at a high rate to try to grow market share but if they aren’t then they’re definitely in the ‘too good to be true’ category.
They basically say they oversubscribe in their marketing materials, but that’s fine for this use case. If I had enough users for that to matter, I wouldn’t be penny pinching.
Trying to setup my 3 rasp pies … Got 1 working so far with pi-hole.
Having a hard time using mkfs on the samsung 1Tb ext hard drive though.
Bought two Chromebooks and have been playing around getting Linux installed but went back to ChromeOS when I realized if I wanted a true Linux machine I need to flash the firmware. So back to the container within the VM. Waiting on some bits to put together a makeshift SuzyQable since the supply chain crisis has made the orange ones hard to come by. Once I have that working (with just a tiny bit of soldering, thank God) I’ll hopefully be able to flash the firmware on one of them so as to have that aforementioned machine. After that, who knows! Maybe give it away to someone or use it as a beater. Otherwise, having not used a Chromebook in years this low-end HP model has been surprisingly pleasant to setup and use and has pretty sweet speakers for those Steve Roach ambient classics of the 1980s.
For work, I’m writing up what our dev team did over the last year to give perspective hires some info on what our worklife is like. I work for Norway’s biggest subscription newspaper, and our big highlights: redesign of the frontpage to provide users with more context and increase time well spent when browsing us. Work with an ML trained version of our podcast journalists which will provide text-to-speech for all articles. Using AI to generate real estate trend articles. And a stream dedicated to working with innovation in a lean manner. We also spent one day every fortnight working on self-improvement, ranging from ML, Git, React, language development. If that sounds compelling to you, we’re hiring people of all levels in the job add here. Feel free to mention Lobsters in your cover letter.
In my private time, I’m working away on Derw. I have a monthly blog post coming up going over all the major changes, along with some thoughts and reflections on language design. I’m getting very close to the point where I can start my first big project in Derw itself: a homepage.
This is my last week of sabbatical, and it’s back to work next Tuesday.
So I’m still reading “Topology and its Applications.” It’s been slow going with the holidays, but I’m making progress and slowly absorbing it.
I’ve also been reading through “Time Management for System Administrators” when I’m not up for the math in the topology book.
And I’m still working on my OpenGL library and trying to get outside.
Hammering out a spec for a new social semantic markup language: https://gitlab.com/timmc/cavern/-/blob/master/doc/ideas/markup.md
I have the week off from work, so I’m planning to spend a full 40-hour week just working on my social media protocol. Currently I’m focused on a markup language optimized for social media; some people would be fine with plaintext, but there needs to be a markup language otherwise someone’s going to shove HTML in there, and then it’s all over.
Today I revisited my goals and I’m pretty happy with how it all factored out: Core goals, goal implications, strategies, and then the beginnings of the spec.
(The larger project, Cavern, is a social media protocol with privacy and user empowerment as main goals. It uses an interesting store-and-forward design that allows use of fairly low-tech hosting. It’s my bid for quasi-self-hosted social media that’s nevertheless user-friendly.)
I’m migrating a Rust web service from Rocket to actix-web. It seems to be going well but like most big rust migrations it’s been a day of compiler errors. I’ll probably slow down after vacation but hopefully I can finally get it to compile again and ditch Rocket.
Then write a blog post about my pros and cons of Rocket.
I’m curious, what was your reason for migrating? Do you find that you prefer actix to rocket?
If you are interested in maybe too many works on the topic.
https://kevincox.ca/2022/01/01/rocket-review/
And it was posted to lobsters here https://lobste.rs/s/81k1ik
Yes. Although to be honest I’m not even using much of the actix framework. I’m doing a bunch of stuff myself.
I guess read the blog post for the details… but my main problem with Rocket is that forcing FromRequest usage doesn’t work for me. It is cute and convenient for small examples but it is too limited for a lot of stuff. For example I wanted to add ETag support to a bunch of pages but you can’t do this from a FromRequest implemention (you can only return a status code, not modify the response or generate a custom response) so I would need to do it in the method body, but without being able to have a Request object each of these “middlewares” needs to have its own set of arguments pulled from the request which is a very awkward API. I find actix-web with user-accessible Request and Response objects to be far more compostable. Maybe I’m missing how you are supposed to implement something like this but I think it comes down to using regular code is always going to give you more flexibility than some small FromRequest API where the only per-handler config is at the type level.
I will miss the URL generation from Rocket though. It was really nice to have type-safe link generation but I’m manually making a rough solution for now which is good enough, maybe I’ll try to push something upstream to actix at some point.
Deeply enjoying the book “Word by Word” by Kory Stamper. Best-written (at the word level) book I’ve ever read; unsurprising, because the author is a dictionary writer (sorry, lexicographer). Also, getting ready to perhaps hear a lot about antisemitism in Poland for some stuff I’m doing. Maybe it’ll expand my horizons.
I’m finally getting around to finishing the painting of the walls in my house. Other than that, I intend to get started with stronglifts 5x5 and generally recharge and relax so I can get back to work refreshed in the new year. And perhaps I’ll take some time to work on CHICKEN as well.
Not much in terms of $work.
Working on blog posts listing best SFF books I read this year, reviewing a technical book I’ve been reading this month, may be a post reviewing my own work this year, etc.
I made the launchd automatically update the https://plan.cat/~hauleth on change.
So as it is said in the
.plan- finishing all tasks as last week I learned that my current contract will not be extended, so I will be looking for new contract in new year. After that I need to refresh my CV and start looking for new workplace.I’d like to teach myself some basic GC algorithms: refcounting, full collection, generational collection, etc.
Anyone have any good resources to self-teach these? I can look up discussions of them, I suppose I’m looking for a bit more exposition than commented code.
I strongly recommend https://gchandbook.org/
Thank you, this book looks like exactly what I’m after!