Accessibility
A fair world needs to account for accessibility throughout society. We aren’t perfect, but we can make things better. We should make reasonable accommodations before we are forced to do something.
Imagine a world where every hotel had no braille on doors, water fountains, restrooms, or elevator buttons. Every hotel thinks that braille is not necessary. Every manager says that blindness is extremely rare. “Besides,” they would say, “blind people should only travel with sighted chaperones.” No hotel would take action on braille until the first time that it hosted a convention for the blind. In this world, no hotels would have braille because most blind people would avoid hotels that dismissed or infantilized them. They would not have conventions at a hotel. Nothing would change. The situation would never improve.
When people with disabilities stay out of public life, people make assumptions. They wrongly believe that a person with a disability is rarer than a non-aero drop bar or an Australian drop bear.
We tend to overestimate the cost of accommodating someone with a disability. As a result, an accessibility feature can seem like something extra. It is nice to have but not essential.
In reality, most provisions are not expensive luxuries. For example, when the factory makes the buttons for an elevator, it is easy to change the production machines so that they punch in the braille bumps on each button. Additionally, the bumps can help sighted people to feel smooth buttons in low-light situations. Therefore, the braille bumps help everybody and hurt nobody. We call these benefits the curb cut effect.
The Curb Cut Effect
Accessibility improvements have a huge impact for some people. For everyone else, these changes have the so-called curb cut effect.
That term gets its name from the original curb cut—the part of the sidewalk curb that is a ramp instead of a cliff’s edge. This cutout is vital for the person who uses a wheelchair, but the effect is to help everyone who uses that sidewalk.
Picture the entrance to a building that is a little higher than the sidewalk. The traditional plan would put the entrance one step up from the sidewalk. Instead of the step, however, imagine a ramp.
The ramp makes it possible for someone with a wheelchair to independently enter or exit the building, but other people also reap the benefits. Think of the worker moving heavy furniture into a store, the father pushing a stroller into a medical office, and the injured woman using crutches as she visits a friend’s home.
That is the curb cut effect. Everyone benefits.
The curb cut effect applies elsewhere too. A voice-over or audio narration is necessary for those with vision disabilities, People who can read unassisted will be able to get through books and articles while their hands are not free. They can also listen in when walking or, yes, while washing dishes.
The Deaf will use transcripts for podcast episodes. The ideal is to have live interpretation in the sign language of that place. Just as the hearing use tone of voice, the Deaf do more than replace written words with handmade symbols. Logistically, this is possible for certain state events, and especially with high-definition video. Transcripts aren’t the same experience, but they make consumption possible for the Deaf and provide other benefits as well.
Those who are learning the target language will appreciate transcripts during podcasts or during TV news. Also, hearing is not all or nothing. Damage can be progressive and hearing can be partial. A transcript is like anything: It’s better to have it and not need it than to need it and not have it. Transcripts and soft subtitles can also
Affordability
Just as we need to recognize disabilities, we must be mindful of economic class. The Yellow Light strives to present content in the way that costs the least in terms of energy and equipment. At the same time, we want to maintain the minimum of quality. The World Wide Web should be for all the people in the world, not just for the richest countries or well-off people, and their experiences should be good and not just “good enough.”
The Yellow Light does not want to impose extra costs on followers. Expenses come in the form of patent fees, high-speed connections, energy, and more expensive devices. If someone distributes content in large and complex files, then the user must have a robust internet connection and a personal smartphone or a newer computer. If the file is too complex or requires conversion, then that file will more quickly drain a device’s batteries.
Compatibility
- Alt text
Every image will have alternative text to describe it. - Transcripts
Every episode of the podcast has the dialog written out in a VTT file that notes the speaker at each segment. Transcripts are important because they open the door for people with hearing loss and assist English learners. The embedded player from Public Radio Exchange (PRX) displays the transcript and speaker labels, as do the Apple Podcasts app, Antenna Pod (Google Play and F-Droid), and Pocket Casts (App Store and Google Play). - Voice-overs
You can listen to every article. You can play a voice-over directly through Ghost’s embedded player on the website or download the file. Because you are listening to a human, your brain has an easier time with processing and remembering information. The positive effects hold true even when you cannot readily tell if a voice is human or AI. - Loudness
Every audio and video file from The Yellow Light has the same perceptual loudness. Listeners will not have to adjust the volume controls back and forth when using the embedded players or when playing back downloads on a program that cannot normalize the volume. The files use the standard loudness of most podcasts: -16 dB LUFS. - Audio formats
In order to accommodate slower connection speeds, use less energy, play on any almost any device, allow accurate seeking, and remain clear, the podcasts are in plain MP3 files, stereo, at a constant bit-rate of 128 kbps, with a sampling rate of 44.1 kHz. Voice-overs are mono and 64 kbps in mono at 48 kHz—the same as the voice training data for Mozilla Common Voice. - Document formats
Files with text will not be in Microsoft- or Apple-only formats. PDF files are the solution. They look the same everywhere, run anywhere, and print easily. Plus, you can open them with the simplest of programs on the cheapest of computers. Plain-text files are in Markdown, end in “.txt,” use the <CR><LF> line ending, and end with a blank line—for maximum compatibility. Every language uses letters and punctuation not in the basic character set. Of the various standards for text encoding, UTF-8 (Unicode) is the most inclusive; all competitors have fallen by the wayside. - Image formats
Pictures files will be GIF, JPEG, or PNG. - Tags
Every voice-over file and podcast episode has an embedded ID3 tag. This ensures that the cover art and almost all of the information from a podcast app stays with the file after download. To be compatible with almost everything after February 1999, the ID3 tags are version 2.3, not 2.4, which never gained traction. The image within each tag measures 512 by 512 pixels as PNG or JPEG (8-bit RGB without transparency or progressive optimization). The entire tag is under 200 kB in size. - Organization
Every podcast episode will have notes in the ID3 tag for the title of the podcast (artist, album), author (album artist), and copyright. In each episode in the RSS feed and ID3 tag, are the item’s title, the episode number (track), and show notes (comment). When a podcast series is episodic and not serialized, no episode will have a season (“disc”) number.
Transcript Policy
The Yellow Light’s transcripts are in soft form. Such files stand in contrast to so-called hard transcripts and subtitles that are “burned in.” The video file itself will have the words etched into the picture. This style of captioning poses several issues.
- Quality
If you want to burn in the subtitles and keep the video file at the same size, then you have to lower the quality of the video. If you want to keep the video’s quality the same with the burned-in subtitles, then you need to make a bigger file or use a more complex format. A bigger file costs more energy to create and to send. A more complex format uses more energy to create the file and to play it. - Showing or hiding
Moving text and colorful effects can be distracting for some users. No user, however, can change anything if the author burned the subtitles into the video. - Editing
With soft subtitles, you have a separate subtitle file in order to translate the dialog or to fix mistakes. The author can make things easier on English learners by fixing typos, misspellings, and missing words. - Font
When users have soft transcripts or subtitles, they can change the look of the transcripts or subtitles. People with low vision may need to set different colors. Someone who has dyslexia may prefer special thickness and shapes. With soft subtitles, the user may be able to use a special font to achieve that effect.
Hosting
The Yellow Light holds the domain (theyellowlight.org) for the website, the RSS feed, and the email. This also means the exclusive right to self-verify at this location for Bluesky.
Control of the domain is separate from the hosting for the email, the web pages and media files. Because The Yellow Light can control the domain separately, hosting providers can change while every link, bookmark, and email address will still work.
This situation allows The Yellow Light to use any hosting provider and change providers in the future. It is important to choose hosts that respect accessibility and compatibility. The Yellow Light has the freedom to use various hosting providers that fulfill the requirements.
The hosting provider for the main website had to allow embedded players and file attachments. Beehiiv was not an option because it blocks most file-types. The hosting provider must also allow downloads of voice-over files. Substack, for example, does not allow downloads for voice-overs.
For the podcast, it was necessary that the hosting provider allow the addition of transcripts. Some hosting providers charge extra money for speech-to-text transcription with time-stamps. As a result, The Yellow Light uses TurboScribe, which is free of charge for files under thirty minutes.
As another requirement, the hosting provider should not re-encode sound files. Substack and SoundCloud re-encode audio and are therefore unacceptable.
Encoding Once and Only Once
The Yellow Light encodes audio files with certain consistent specifications. This process of encoding can compress the file by more than 90%. It is a “lossy” compression, but with the right encoding parameters, you can do a single lossy compression without any human being able to hear the difference. Encoding a second time introduces many problems, however.
Some hosts will encode all audio files a second. This is called re-encoding, If you re-encode a file that was already encoded, then that is as if you took a blurry picture of a photocopy. A photograph is fine. A photocopy is also fine. A photograph of a photocopy is less fine.
If a hosting provider re-encodes the audio files, then it degrades the quality of the file and destroys the ID3 tags. The Yellow Light puts an ID3 tag into every audio file so that the file still has its labels after being downloaded and moved to another app. Without the ID3 tag, the audio file loses its markings.
The hosting provider that re-encodes content will also rename the files. If a downloaded file has no ID3 tag and the original file-name has changed, then you cannot tell which files are what.
Easy Operation on Every Device
The Yellow Light takes great care to make the audio files sport a virtually universal compatibility. Besides the files themselves, the choice of the hosting service had to mind that the host did not tamper with the files.
If a website automatically re-encodes a file, then the file may also have its sampling rate change. The only issue is the act of distortion during the resampling from a re-sampler that users cannot inspect.
Almost all equipment uses the sampling rate of 48 kHz. The rate of 44.1 kHz is just as good and can capture all sounds that humans can hear, and although it is rarely a native sampling rate of any device, this holdover of the CD era hangs on. At one time, audio-books commonly found their way to so-called MP3 CDs. Most podcasts also stick to 44.1 kHz out of tradition.
Another problem comes from redoing lossy compression if the hosting site performs a re-encoding. With a file in a format such as MP3, the sound went through one round of lossy compression. You can go through one round of this and never be able to detect a difference with your ears. The story is different when it comes to doing lossy compression on a file that already went through lossy compression. Imagine a photocopy of a photocopy, a JPEG of a JPEG, or a still image from a video camera’s footage of the printout of a smartphone camera’s photograph of a computer monitor.
The ridiculousness of this workflow does not change even if you use the best smartphone on the market. Likewise, even the best software in the world cannot change the fact that you should never re-encode a file that is already encoded with lossy compression.
SoundCloud does this re-encoding when users stream but not when they download. If you upload an MP3 to SoundCloud, then the downloadable version will be fine, but the streaming version will be an MP3 of an MP3. If you upload an original copy of your audio, then the streaming version will only do lossy compression once, but the downloadable version will be a huge file.
Substack always re-encodes your uploads. This seems better than SoundCloud’s policy, but with either host, you lose control over the tags, the filenames, the quality, and so on. To make matters worse, Substack will choose the bitrate differently each time and make decisions over mono versus stereo. You only find out after the fact, and no one can predict what will happen with an uploaded file.