personal comments edit

His Royal Highness, the High Commander Staniel James Wonkfoot, Professor of Defense Against the Dark Arts

On Saturday, November 14, 2026, we had to say goodbye to our dear little man, Stan. He was 18-1/2 years old, which is a good long life for a cat.

We brought Stan and his brother, Kai, home in July 2008 as kittens, brothers from the same litter. Kai was always a little bigger than Stan, but Stan ran the show and we all knew it.

Kai and Stan, brothers. Kai has a gray nose, Stan has a pink nose

With our little girl cat, Xev, we named them after one of our favorite shows, Lexx.

Stan, Kai, and Xev - all in a cat pile

Originally his name was from the show - Stanley Tweedle. However over the course of his life he became His Royal Highness, the High Commander Staniel James Wonkfoot, Professor of Defense Against the Dark Arts.

Yeah, that’s long. He earned it.

He was so demanding and persistent. An alpha, leading us around, herding us, and absolutely telling us about all of his problems. Very loudly. Repeatedly.

He lived through a remodel and having to move out of the house twice for our wood floors getting refinished. One of those times he lived with us in a pet-friendly long term stay hotel for a couple weeks. He was a total trooper. Never had a single accident.

As he got older, he decided one day he just wasn’t going to go upstairs to where his food, water, and box were. Just, nope. We figured it out pretty quickly and moved things down for him. He mostly stayed downstairs after that, his arthritis giving him trouble.

He became my work buddy and I put a bed in my office by my desk so he could supervise. He liked to wake up, stretch, and then scratch on my mesh garbage can to get my attention, calling for pets. I’d pet him, maybe sing him a song, maybe carry him like a baby to our back glass door and sing him rockabye. He loved the attention.

He really loved his kisses. He’d do this thing where he’d walk across my chest so his front half was on me but his back half was on a pillow next to me, then he’d “slow flop” - the front half would flop down, then the back half would take a few seconds and follow. The head butts would commence and he wanted kisses on his head. Scratch the chin while you’re doing that, too, human.

We found a lump in his tail and one on his side a couple years back. Cancer. He was already pretty old so we figured we’d let him live Best Cat Life instead of putting him through painful treatments that would make him miserable. He’d tell us when it was time to go.

I really wanted him to fall asleep and go peacefully, but instead we woke up on Saturday and he was having repeated seizures. The cancer had finally gotten to his brain. We took him to the vet and said goodbye.

This is the first time in… a long time. Over 20 years? That we haven’t had a cat around here. I don’t have anyone supervising my work efforts or yelling at me about how I’m not sitting down fast enough or giving the appropriate number of kisses. I don’t have anyone slapping my leg to get my attention or grabbing my hand to pull it over and get pets. GTA6 comes out tomorrow. He won’t be playing with me. I don’t think I’ve ever played a GTA game without a Game Cat to keep me company. Stan was older than Phoenix. She’s almost 16. She’s never known a life without him. She’s not getting her morning cuddles.

We’re all pretty lonely without our little man. I just hope that, wherever he is, he’s found Xev and Kai and he’s having adventures with them now.

Goodbye, my good friend. We love you so much.

Stan 'slow-flopped' on me and fell asleep on my arm - November 12, 2026

halloween, costumes comments edit

This year we had 317 trick-or-treaters. That’s our new all-time record, at least since I’ve been tracking.

Trick-or-Treaters by Time2019 (190)2021 (140)2024 (235)2025 (227)2026 (317)02550751005:30 - 6:006:00 - 6:306:30 - 7:007:00 - 7:307:30 - 8:008:00 - 8:302019, 6:00 - 6:30: 12019, 6:30 - 7:00: 562019, 7:00 - 7:30: 592019, 7:30 - 8:00: 412019, 8:00 - 8:30: 332021, 6:00 - 6:30: 162021, 6:30 - 7:00: 372021, 7:00 - 7:30: 302021, 7:30 - 8:00: 502021, 8:00 - 8:30: 72024, 5:30 - 6:00: 242024, 6:00 - 6:30: 222024, 6:30 - 7:00: 822024, 7:00 - 7:30: 522024, 7:30 - 8:00: 262024, 8:00 - 8:30: 292025, 5:30 - 6:00: 382025, 6:00 - 6:30: 572025, 6:30 - 7:00: 412025, 7:00 - 7:30: 432025, 7:30 - 8:00: 442025, 8:00 - 8:30: 42026, 6:00 - 6:30: 662026, 6:30 - 7:00: 832026, 7:00 - 7:30: 582026, 7:30 - 8:00: 562026, 8:00 - 8:30: 546683585654
Average by Time Block02040605:30 - 6:00: 31.0315:30 - 6:006:00 - 6:30: 24.1246:00 - 6:306:30 - 7:00: 59.0596:30 - 7:007:00 - 7:30: 57.3577:00 - 7:307:30 - 8:00: 44.3447:30 - 8:008:00 - 8:30: 16.5178:00 - 8:30
Year-Over-Year Trick-or-Treaters010020030040020052006200720082009201020112013201420152016201920212024202520262005: 1551552006: 1621622007: 1391392008: 2372372009: 2432432010: 2592592011: 1891892013: 2982982014: 1761762015: 85852016: 1841842019: 1901902021: 1401402024: 2352352025: 2272272026: 317317

Halloween was on a Saturday and it was nice and clear.

We forgot we were starting at 5:30 and we were running late anyway, trying to eat dinner, so we didn’t start until 6:00. We did have a few doorbells between 5:30 and 6:00, but not too many, so I can’t imagine we really missed a lot. It did seem like 5:30 was pretty early to get started.

Missing/incongruous data:

  • 2017 we were remodeling the house and didn’t hand out candy.
  • 2018 I walked with Phoenix and the tally got lost somewhere.
  • 2020 was COVID lockdown and no one handed out candy.
  • 2022 we were getting our floors redone due to a leak and the whole house was torn up so we didn’t hand out candy.
  • 2023 we had COVID and were too sick to hand out candy.

Cumulative data:

 Time Block
Year5:30 - 6:006:00 - 6:306:30 - 7:007:00 - 7:307:30 - 8:008:00 - 8:30Total
2005155
2006525935160162
2007549392521139
20081471824525237
20091751728221243
20101977764839259
2011318053250189
20132872113805298
20141954514210176
201513143028085
201615967570184
2019156594133190
2021163730507140
2024242282522629235
202538574143444227
20266683585654317

Our costumes this year:

  • Me: Ghostbuster
  • Jenn: Joker
  • Phoenix: (Not pictured) Coraline

Jenn as Joker, me as a Ghostbuster

I’ve wanted to be a Ghostbuster since I can remember. When the HasLabs proton pack and wand came out, I ordered it and I’ve been planning this for almost a year.

I didn’t sew any of the stuff this time. Since COVID, apparel fabric shops are scarce. Plus, for a flight/jumpsuit like this, the amount of time and materials to make it vs. build it doesn’t make sense. Vastly cheaper to buy it.

I did modify the pack to have a larger speaker so it sounds better. I also added some aftermarket stickers to it that stick far better and look more realistic than the stock stickers.

Me as a Ghostbuster

What’s cool is that it also goes with our outdoor decor since we have a Stay-Puft Marshmallow Man and Slimer on our front porch.

It's the Stay-Puft Marshmallow Man

I would occasionally go outside, “power up” the pack (which makes all the noises like it’s spinning up), and “shoot” the proton wand at the kids up and down the street. This, admittedly, seemed to draw them in as some would cross the street to come over after this. I can’t say this didn’t contribute to the increased number of trick-or-treaters. I had one trick-or-treater ask if they could get a picture of me by Stay-Puft and I was happy to oblige.

Parents also appreciated the costume and we found several groups singing the Ghostbusters theme song or yelling, “WHO YOU GONNA CALL?” from the sidewalk.

I will likely reuse this many times over coming years since it’s an easy costume to get into, it’s reasonably comfortable, and it’s a big hit with the kids.

This year was also, unfortunately, the last year for the annual Halloween party at our good friends’ house that we’ve gone to for… well, this is 26 years, I think.

I enjoy sewing and making costumes, for sure. It’s gotten a lot harder with the lack of fabric stores, especially now that Jo-Ann is gone. Lots of quilting fabric, very little apparel fabric. I was making a new costume every year to start so I could be something different each year at that Halloween party and at work in the office. After that, it was also so I could go walking around with Phoenix.

My kiddo is getting too old to walk around the neighborhood. COVID shut down my office, and I think contributed to the loss of Jo-Ann. Now that the Halloween party is over… it’s just me.

Which won’t stop me from making things, of course, but it just means I can be more thoughtful about what I want to make and also reuse the costumes I make instead of going for something new every year.

But, dammit, I’ll sure miss those parties.

home, javascript, maker comments edit

I built a little in-browser app that can create a customizable, full-color 3D model of a coaster with a QR code on it, ready for printing. Go check it out!

There are a ton of these apps out there - apps that create QR codes and let you export them or put them on a 3D printable coaster, but I found there were problems with all of these:

  • They’re not free. You want to export an SVG? Pay. You want to customize the logo in it? Pay. It seems like creating a QR code shouldn’t cost you.
  • They require an account. Sign in before you finish! I’m not sure why. I think this is one of those “if you don’t know what the product is, you’re the product” things.
  • They send stuff to a server. I do not want to send my guest wi-fi password to your server. I do not want to send my contact info to your server. I just want a QR code.
  • They’re not feature-complete. They don’t have the options I want. Maybe the model is sub-par. Maybe getting a color print model is hard or manual. In most cases the 3D print part seems like a “bolt-on” instead of a real feature.

I decided to fix it myself.

QR Code Coaster generator screenshot

Key differentiating features:

  • No server connection. All the work happens in your browser. The rendering, the testing, the generation of the downloadable model. Your information never leaves your machine. That also means there’s no account required, no payment required, it all just works.
  • 3D printing first. The goal is a 3D printable coaster, not just a QR code generator that happens to also create a model.
    • Print and scan testing. Simulated tests will run based on your selected nozzle size to see if you’ll be losing details or if you’ll have problems scanning the code after print. This allows you to correct the issues before finding out on the printer.
    • Fixed color management. Choose up to four filament colors and ensure all colors used in the model match those colors.
    • Nice edges. The top edge can be slightly rounded with the bottom edge at a nice fillet (no supports!) to make the edges nice instead of sharp.
  • Highly customizable. Select from tons of options to make the coaster yours.
    • Type and style of QR code.
    • Logo/icon in the center of the QR code.
    • Optional text above/below the code.

Once it looks how you want, you can download a .3mf file with the geometry, pre-colored using your palette.

I, personally, am a Bambu X1C owner, so I pop this open using Bambu Studio, match the colors to colors in my AMS, and I’m off to the races.

Give it a try! If you want to see more, the source code is here.

Back in 2017 I created an external volume monitor for our Marantz receiver. The receiver is in a cabinet so we can’t see the display on the front, and the on-screen display doesn’t work over 4K inputs. Rather than guess at the volume (to know if you’re going to be blasted out or not be able to hear), I created a little monitor display with an Arduino that kinda did the job.

This is what the original version looked like:

Original volume monitor

At the time I was pretty good with it. It solved the problem and was… fine.

It’s nine years later and over time I got frustrated with some of the shortcomings. The contrast on the display was never super, so it was always just a little difficult to read from a distance. Combined with my eyes (getting older), it wasn’t cutting it anymore. After a power outage, it was a pain to “recover” because you had to ensure the Arduino was powered up after the receiver or it’d go into a death loop.

Hardware has progressed a lot, so I decided to make an upgrade.

Presenting the new version:

Updated volume monitor

This version has some pretty major upgrades. Built on an ESP32 (instead of Arduino) with a 4” touch screen (instead of a 1602 display), it now has:

  • Wi-Fi support instead of wired networking.
  • Auto-detection of networks.
  • Auto-detection of receivers on the network.
  • A nice, large volume display along with a visual gauge.
  • Improved display of the source and sound format.
  • Full settings menu for reconfiguration, calibration, or checking current connection information.

All still on a budget - the hardware cost about $30.

Here’s the first page of the settings menu:

Settings menu

(Admittedly, the pictures aren’t great… but it’s hard to take a good picture of a reflective touch screen.)

I’ve updated the GitHub repo where all the source is if you want to build one, too. It has all the software, the hardware list, wiring diagrams, even the 3D model I made for the box.

autofac, dotnet comments edit

Yesterday I released Autofac 9.1.0. It’s listed as a minor release - 0.X.0 semver increment - because it’s all new, backwards-compatible features, but it’s actually pretty big, I think.

AnyKey Support

Autofac now natively supports the concept of AnyKey. It behaves the same way AnyKey works in Microsoft.Extensions.DependencyInjection, but it is native to Autofac directly. The unit tests here show some very detailed examples of usage, but on a high level:

var builder = new ContainerBuilder();
builder.RegisterType<Service>().Keyed<IService>(KeyedService.AnyKey);

var container = builder.Build();

// Registering as AnyKey allows it to respond to... any key!
var service = container.ResolveKeyed<IService>("service1");

As noted, it behaves like AnyKey in Microsoft.Extensions.DependencyInjection, so there are some “rules” around when AnyKey will work and when it won’t.

  • You can resolve a single instance and AnyKey will be a fallback. If you resolve a single instance of a service and use a key that you haven’t otherwise registered, an AnyKey service can provide that instance.
  • You can resolve a collection using AnyKey. This allows you to get all keyed services of a certain type, but it will not include the services registered with AnyKey - only services registered with a specific key.
  • AnyKey works with singletons. When you register an AnyKey service as a singleton, a different instance of that service will be cached for each key under which it’s resolved. Beware of this - if you’re not paying attention, it could lead to a memory leak.
  • Registration order is important. As always, last-in-wins. If you register a lot of AnyKey services, the last one registered for the given type will be the one you get when you resolve.

Here are some examples showing usage:

var builder = new ContainerBuilder();
builder.RegisterType<Service1>().Keyed<IService>(KeyedService.AnyKey);
builder.RegisterType<Service2>().Keyed<IService>("a");
builder.RegisterType<Service3>().Keyed<IService>("b");
var container = builder.Build();

// This will ONLY get Service2 and Service3 - things registered with explicit
// keys. It will NOT return Service1, registered with AnyKey.
var allServices = container.ResolveKeyed<IEnumerable<IService>>(KeyedService.AnyKey)

// THIS WILL THROW: You can't resolve a single instance using AnyKey.
_ = container.ResolveKeyed<IService>(KeyedService.AnyKey);

// This will get you the AnyKey service because no specific keyed service was
// registered with `other-key`.
var noRegisteredKey = container.ResolveKeyed<IService>("other-key");

Again, check out the unit tests for some robust examples.

Inject Service Key Into Constructors

The new [ServiceKey] attribute allows you to inject the service key provided during resolution. This is handy in conjunction with AnyKey and is also similar to the construct in Microsoft.Extensions.DependencyInjection, but with native Autofac.

First, mark up your class to take the constructor parameter.

public class Service : IService
{
  private readonly string _id;
  public Service([ServiceKey] string id) => _id = id;
}

Then when you resolve the class, the service key will automatically be injected.

You can also make use of this in a lambda registration.

var builder = new ContainerBuilder();
builder.Register<Service>((ctx, p) => {
  var key = p.TryGetKeyedServiceKey(out string value) ? value : null;
  return new Service(key);
}).Keyed<Service>(KeyedService.AnyKey);

Metrics

Some metrics have been introduced that can allow you to capture counters on how long middleware is taking, how often lock contention occurs, and so on.

Set the AUTOFAC_METRICS environment variable in your process to true or 1 to enable this feature. You can see the set of counters that will become available here.

⚠️ This is NOT FREE. Collecting counters and metrics will incur a performance hit, so it’s not something you want to leave on in production.

General Performance Improvements

A pass over the whole system was made with an eye to trying to claw back some performance. Some additional caching was introduced to help reduce lookups and calculations of reflection data; and a few hot-path optimizations were made for the common situations.

A Personal Note

This is probably the biggest Autofac update made in quite some time - new features, some perf fixes, some metrics - and, with that, I really don’t have any active assisting project maintainers, so it was all just me. I’m pretty proud of this one. I have been suffering from maintainer burnout for a while, but I’ve got a little of my energy back lately and, honestly, it’s due to AI.

I used Codex and Claude Opus on this latest set of changes to help me out and do some of the heavy lifting. Don’t get me wrong, I reviewed all of what was output, but it was so energizing to be able to tell something else to go look for changes, dig for performance optimization opportunities, and validate them. I could create a robust set of tests (or, in some cases, adapt existing Microsoft container compatibility tests) to guide implementation and basically say, “Make it pass the tests.” I can safely say I would not have been able to deliver this new set of functionality in any near future without tools like that. Between time constraints and flagging motivation, it wouldn’t have happened.

I’m not at “just vibe code everything” level and I’m definitely not ready for Gas Town, but I see the value here and it counters some of the burnout.