This post contains some brief notes about building offline, static web sites using React, in order to further the objectives of minimal computing. But before I go there, first let me give you a little background…

The Lakeland Community Heritage Project is an effort to collect, preserve, and interpret the heritage and history of African Americans who have lived in the Lakeland community of Prince George’s County, Maryland since the late 19th century. This effort has been led by members of the Lakeland community, with help from students from the University of Maryland working with Professor Mary Sies. As part of the work they’ve collected photographs, maps, deeds, and oral histories and published them in an Omeka instance at As Mary is wrapping up the UMD side of the project she has become increasingly interested in making these resources available and useful to the community of Lakeland, rather than leaving them embedded in a software application that is running on servers owned by UMD.


Recently MITH has been in conversation with LCHP to help explore ways that this data stored in Omeka could be meaningfully transferred to the Lakeland community. This has involved first getting the Omeka site back online, since it partially fell offline as the result of some infrastructure migrations at UMD. We also have been collecting and inventorying disk drives of content used by the students as they have collected and transfer devices over the years.

One relatively small experiment I tried recently was to extract all the images and their metadata from Omeka to create a very simple visual display of the images that could run in a browser without an Internet connection. The point was to provide a generous interface from which community members attending a meeting could browse content quickly and potentially take it away with them. Since this meeting was in a environment where there wasn’t stable network access it was important that for the content to be browsed without an Internet connection. We also wanted to be able to put the application on a thumb drive, and move it around as a zip file, which could also ultimately allow us to make it available to community members independent of the files needing to be kept online on the Internet at a particular location. Basically we wanted the site to be on the Sneakernet instead of the Internet.

Static Data

The first step was getting all the data out of Omeka. This was a simple matter with Omeka’s very clean, straightforward and well documented REST API. Unfortunately, LCHP was running an older version of Omeka (v1.3.1) that needed to be upgraded to 2.x before the API was available. The upgrade process itself leapfrogged a bunch of versions so I wasn’t surprised to run into a small snag, which I was fortunately able to fix myself (go team open source).

I wrote a small utility named nyaraka that talks to Omeka and downloads all the items (metadata and files) as well as the collections they are a part of, and places them on the filesystem. This was a fairly straightforward process because Omeka’s database ensures the one-to-many-relationships between a site and its collections, items, and files which means they can be written to the filesystem in a structured way:

This post was really meant to be about building a static site with React, and not about extracting data from Omeka. But this filesystem data is kinda like a static site, right? It was really just laying the foundation for the next step of building the static site application, since I didn’t really want to keep downloading content from the API as I was developing the application. Having all the content local made it easier to introspect with command line tools like grep, find and jq as I was building the static site.


Before I get into a few of the details here’s a short video that shows what the finished static site looked like:

Lakeland Static Site Demo from Ed Summers on Vimeo.

You can see that content is loaded dynamically as the user scrolls down the page. Lots of content is presented at once in random orderings each time to encourage serendipitous connections between items. Items can also be filtered based on type (buildings, people and documents). If you want to check it out for yourself download and unzip this zip file and open up the index.html in the directory that is created. Go ahead and turn off your wi-fi connection so you can see it working without an Internet connection.

When building static sites in the past I’ve often reached for Jekyll but this time I was interested in putting together a small client side application that could be run offline. This shouldn’t be seen as an either/or situation: it would be quite natural to create a static site using Jekyll that embeds a React application within it. But for the sake of experimentation I wanted to see how far I could go just using React.

Ever since I first saw Twitter’s personal archive download (aka Grailbird) I’ve been thinking about the potential of offline web applications to function as little time capsules for web content that can live independently of the Internet. Grailbird lets you view your Twitter content offline in a dynamic web application where you can view your tweets over time. Over the past few years the minimal computing movement has been gaining traction in the digital humanities community, as a way to ethically and sustainably deliver web content without needing to make promises about keeping it online forever, or 25 years (whichever comes first).

React seemed like a natural fit because I’ve been using it for the past year on another project. React offers a rich ecosystem of tools, plugins and libraries like Redux for building complex client side apps. The downside of using React is that it is not as easy for people to set up out of the box, or for changing over time if you you aren’t an experienced software developer. With Jekyll it’s not simple, but at least its relatively easy to dive in and edit HTML and CSS. But on the plus side for React, if you really want to deliver an unchanging, finished (static) artifact, then maybe these things don’t really matter so much?

At any rate it seemed like a worthwhile experiment. So here are a few tidbits I learned when bending React to the purposes of minimal computing:

Static Database

The first is to build a static representation of your data. Many React applications rely on an external REST API being available. This type of dependency is an obvious no-no for minimal computing applications, because an Internet connection is needed, and someone needs to keep the REST API service up and running constantly, which is infrastructure and costs money.

One way of getting around this is to take all the structured data your application needs and bundle it up as a single file. You can see the one I created for my application here. As you can see it contains metadata for all the photographs expressed as JSON. But the the JSON itself is part of a global JavaScript variable declaration which allows it to be loaded by the browser without relying on an asynchronous HTTP call. Browsers need to limit the ability of JavaScript to fetch files from the filesystem for security reasons. This JavaScript file is loaded immediately by your web browser when it loads the index.html, and the app can access it globally as window.DATA. Think of it like a static read-only, in memory database for your application. The wrapping HTML will look as simple as something like this:

<!DOCTYPE html>
<html lang="en">
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <title>Lakeland Community Heritage Project</title>
    <script src="static/data.js"></script>
    <div id="app"></div>
    <script type="text/javascript" src="bundle.js"></script>

Update: Another more scalable approach to this suggested by Alex Gil after this post went live, is to try using an in browser database like PouchDB. When combined with Lunr for search this could make for quite a rich and extensible data layer for minimal computing browser apps.

Static Images

Similarly, the image files need to be available locally. I took all the images and saved them into a directory I named static, and named the file using a unique item id (from Omeka) which allowed the metadata and data to be conceptually linked:


My React application has an Image component that simply renders the image along with a caption using the <figure>, <img> <figcaption> elements.

  • image
class Image extends Component {
  render() {
    return (
      <Link to={'/item/' + + '/'}>
        <figure className={style.Image}>
          <img src={'static/' + + '/fullsize.jpg'} />

It’s pretty common to use webpack to build React applications, and the copy-webpack-plugin will handle copying the files from the static directory into the distribution directory during the build.


You may have noticed that in both cases the data.js and images are being loaded using a relative URL (without a leading slash, or a protocol/hostname). This is a small but important detail that allows the application to be moved around from zip file, to thumb drive to disk drive, without needing paths to be rewritten. The images and data are loaded relative to where the index.html was initially loaded from.

In addition many React applications these days use the new History API in modern browsers. This lets your application have what appear to be normal URLs structured with slashes which you can manage with react-router. However slash URLs are problematic in a offline static site for a couple reasons. The first is that there is no server so you can’t tweak it to respond to any request with the HTML file I included above that will bootstrap your application. This means that if you reload a page you will get a 404 not found.

The other problem is that while the History API works fine for an offline application, the relative links to bundle.js, data.js and the images will break because they will be relative to the new URL.

Fortunately there is a simple solution to this: manage the URLs the way we did before the History API, using hash fragments. So instead of:


you’ll have:


This way the browser will look to load static/data.js from file:///lakeland-images/ instead of file://lakeland-images/index.html/items/. Luckily react-router lets you simply import and use createHashHistory in your application initialization and it will write these URLs for you.


It’s important to reiterate that this was an experiment. We don’t know if the LCHP is interested in us developing this approach further. But regardless I thought it was worth just jotting down these notes for others considering similar approaches with React and minimal computing applications.

I’ll just close by saying in some ways it seems counter-intuitive to refer to a React application as an example of minimal computing. As Alex Gil says:

In general we can say that minimal computing is the application of minimalist principles to computing. In reality, though, minimal computing is in the eye of the beholder.

After working with React off and on for a couple years it still seems quite complicated–especially when you throw Redux into the mix. Assembling the boilerplate needed to get started is still tedious, unless you use create-react-app which is a smart way to start. By comparison it’s much easier to get Jekyll out of the box and start using it. But, if the goal is truly to deliver something static and unchanging, then perhaps this up front investment in time is not so significant.

Static sites, thus conceived, ultimately rely on a web browser, which are insanely complicated pieces of code. With a few exceptions (e.g. Flash) browsers have been pretty good at maintaining backwards compatibility as they’ve evolved along with the web. JavaScript is so central to a functioning web it’s difficult to imagine it going away. So really this approach is a bet on the browser and the web remaining viable. Whatever happens to the web and the Internet we can probably rely on some form of browser continuing to exist as functioning software, either natively, or in some sort of emulator, for a good time to come…or at least longer than the typical website is kept online.

Many thanks to Raff Viglianti, Trevor Muñoz and Stephanie Sapienza who helped frame and explore many of the ideas expressed in this post.