lost in tab

(Re)Making Cirrus: Three lessons learned

Wed, 18 Mar 2026 12:00:00 +0000

Reflecting on five-plus years of building Cirrus, Element 84’s data orchestration framework, three core lessons emerge that apply to software projects generally: follow the money to understand what drives development, expect the pendulum of overcorrection and iteration, and hold fast to an unchanging vision. Part 2 of 2.

(Re)Making Cirrus: Reflections on five years spent building an open-source data orchestrator

Mon, 23 Feb 2026 12:00:00 +0000

Cirrus, Element 84’s open-source data orchestration framework, recently turned five years old — and we’ve rebuilt it at least three times along the way. This is a retrospective on where Cirrus came from, from its origins in NASA’s Cumulus through the STAC Workflows idea, the project CLI, the SWOOP detour, and finally the v1.0.0 release, plus a look at where it’s headed next.

Metadata makes the data format: metadata storage and representation across array formats

Tue, 04 Nov 2025 12:00:00 +0000

When configured equivalently, COG and Zarr store the exact same data bytes. So what actually distinguishes one raster format from another? Metadata. We get philosophical about what metadata even is, examine how TIFF and Zarr each represent it, tour the various ways geospatial coordinate information is encoded across formats, ask where metadata should live, and make the case that formal, versioned conventions are what truly enable interoperability.

Redefining cloud native with the Coalesced Chunk Retrieval Protocol

Tue, 23 Sep 2025 12:00:00 +0000

What happens if we question current assumptions around Cloud-Native Geospatial (CNG) and imagine what it could be if we were to redefine it with new technologies? The Coalesced Chunk Retrieval Protocol (CCRP) is one idea born of this thinking, aiming to help make CNG more efficient and easier for both data consumers and producers.

A new look, a new domain, a new start

Sat, 23 Aug 2025 08:00:00 +0000

Back from the dead with a new look and a new domain!

Beyond the default: a modern guide to raster compression

Wed, 06 Aug 2025 12:00:00 +0000

Compression is a fundamental and increasingly critical part of modern data storage and processing. But compression is often naively miscategorized as a single processing step. A better understanding of the steps in compression pipelines, the priciples behind them, and how to measure their effectiveness will enable us to move beyond the default compression settings and realize gains from more effective compression techniques.

Chunks and chunkability: an origin story

Wed, 09 Jul 2025 12:00:00 +0000

Raster data formats are chunked internally, but the cloud has exposed these internal implementation details. We dig into the finer details of chunking to understand how chunking has become something tyrannical, dictating how we can and can’t access data efficiently. Part 2 of 2.

Chunks and chunkability: tyranny of the chunk

Wed, 09 Jul 2025 12:00:00 +0000

Is Zarr the new COG?

Thu, 22 May 2025 12:00:00 +0000

We dig into the Zarr and COG raster formats to see if Zarr can and should now be a replacement for COG, and find a surprising answer.

QEMU services with socat

Fri, 19 Jun 2020 12:00:39 -0700

QEMU Services?

I’ve been working on an integration testing environment for a few interconnected services. Such a thing would be easy with containers, but I am modeling an environment of bare metal hosts running these services, leaving little choice but VMs. It’s been easy testing the VMs in isolation using QEMU, which by default with -nographic drops one into the console.

But when one wants to kick off several VMs at once, it’s much easier if they can be started as services in the background. Any number of service supervisiors can run them, that’s no problem. QEMU even has the -daemonize option to background itself.

Serial with socat

Wed, 20 May 2020 23:00:39 -0700

Introduction

I do a lot of work with network devices, which are often headless and require the use of serial consoles for out-of-band access. My go-to on MacOS and Linux has been screen, but various little issues over the years have led me to consider other options. I tried minicom on the recommendation of a colleague, but that didn’t work for me either.

I’ve struggled to pinpoint my problem with the aforementioned tools; I think I just want something less featurful, something that literally only gives me a tty via serial and doesn’t have anything unnecessary between my terminal and the remote shell. Over the weekend, I randomly happened upon an article about using socat for serial on Hacker News, and thought it might be just what I have been searching for. So I tried out the example and it failed miserably with the socat from Homebrew (MacOS being the OS on my daily driver).

Create a static site with Hugo and GitHub pages

Tue, 11 Feb 2020 12:52:22 -0800

As I mentioned in my first post, much of my resistance to blogging has been a blend of prefectionism and an abhorrence to any chance of having to deal with yet more infrastructure issue. That is, I deal with broken crap at my day job; I really don’t want to have to worry about that for my blog. In that vein, I pursued the simplest practical stack with the quickest startup time, ending up with a static site generated with Hugo hosted as a GitHub page.

Introduction

Mon, 10 Feb 2020 22:12:06 -0800

Hello. I am a software engineer from Portland hoping to spread some knowledge when I have something possibly interesting to share.

This is a new thing for me, technical blogging, so it might take me a little time to get started, but I hope I can start to churn out a regular stream of meaningful content quickly.

Why do this?

I have increasingly been feeling the pressure to start a technical blog. I tend to dig into weird, obscure issues fairly frequently, and have often thought my notes could be of interest and/or help to others. But setting up a blog and writing content takes time, and I just haven’t been able to convince myself to make it happen.