I'm a software developer specializing in accessibility for blind and visually impaired people. From mid-2017 to late 2020, I worked for Microsoft on the Windows accessibility team, though my comments were always my own opinion posted on my own initiative.
I'm legally blind myself; I have enough sight to read a desktop computer screen up close with slightly enlarged fonts, but I often use a screen reader when browsing the Web, doing email, and other tasks that don't involve code. (Don't get me wrong; totally blind people can program, but I never got used to doing it that way myself.)
Of course, there's more to me than visual impairment. When it comes to programming, I'm also particularly interested in the challenges of developing cross-platform applications that integrate smoothly with the host platform. Outside of programming, I love music and love to sing. Pre-COVID, I often did karaoke (of course, I memorize the words). And like many nerds, I read science fiction.
Email: mattcampbell@pobox.com
Twitter: @mw_campbell
Blog: http://mwcampbell.us/blog/
Location: Wichita, Kansas, US
[ my public key: https://keybase.io/mwcampbell; my proof: https://keybase.io/mwcampbell/sigs/9sbxkfOoTyW6C8IiDB50bP_ZWa-V73AfbdmvUlfOKe8 ]
- > Did Microsoft seriously deprecate BitBlt and 2D draw calls?
Very unlikely. Far too many applications depend on those things. It's more likely that they accidentally changed something subtle that happened to break colorForth.
- If you're allowed to say, are you referring to the Windows 10 ports of the iOS apps that were done via Osmeta in 2016, or the earlier WinRT-native version? If the former, that was a non-starter for me and my blind friends due to deep accessibility issues, probably having to do with the Osmeta port/reimplementation of UIKit. Edit to add: And we wanted something that was easier to use with a Windows screen reader than the desktop website, particularly for Facebook proper.
- > I still like it -- for systems programming, that is.
Just curious, what language(s) do you prefer for things that you don't classify as "systems programming"?
- > the distinction is whether or not something enables me to perform a task, or whether it's just doing the task for me.
I think school has taught us to believe that if we're assigned a task, and we take a shortcut to avoid doing the task ourselves, that's wrong. And yes, when the purpose is to learn the task or the underlying concepts, that's probably true. But in a job environment, the employer presumably only cares that the task got done in the most efficient way possible.
Edit to add: When configuring or using a particular program is tedious and/or difficult enough that you feel the need to turn to an LLM for help, I think it's an indication that a better program is needed. Having an LLM configure or operate a computer program for you is kind of like having a robot operate a computer UI that was designed for humans, as opposed to having a higher-level program just do the higher-level automation directly. In the specific case of the Apache HTTP Server, depending on what you need to do, you may find that Caddy is easy enough that you can configure it yourself without requiring the LLM. For common web server scenarios, a Caddyfile is very short, much shorter than a typical Apache or nginx configuration.
- > there is no capital or material cost to push back against complexity
On a thread about software bloat and inefficiency, @josephg once speculated about an alternate universe where Moore's Law stopped decades ago. I've kept thinking about that. Unfortunately, I kept coming up with counterfactuals where important things like accessibility weren't as advanced or (relatively) widespread as they are in this world.
- Shouldn't that make us want to fight to simplify our software stacks to the point where we can do analysis by synthesis, building from simple, well-understood parts, again?
- Are you saying that frameworks might become less important because LLMs can just generate boilerplate code instead? Or do I misunderstand? Personally, if the vibe-engineering future that some executives are trying to foist on us means that I'll be reading more code than I write directly, then I want that code to be _doubly_ succinct.
- This may not be what you're after, but note that egui and Slint have accessibility support (at differing levels of completeness), e.g. for blind people using screen readers, while Ribir and GPUI do not.
- What is Sprite in this context?
- The thing that always worries me about these clean-slate designs is the fear that they'll ignore accessibility for disabled people, e.g. blind people, and then either the system will remain inaccessible, or accessibility will have to be retrofitted later.
- In the spirit of "every non-trivial program will expand until ...", I think preemptively choosing async for anything much more complex than a throwaway script might be justified. In this case, the relevant thing isn't performance or expected number of concurrent users/connections, but whether the program is likely to become or include a non-trivial state machine. My primary influence on this topic is this post from @sunshowers: https://sunshowers.io/posts/nextest-and-tokio/
- > They’re trained to be far too user led. They don’t challenge you enough.
An anecdote here: I recently had a conversation with Claude that could be considered therapy or at least therapy-adjacent. To Anthropic's credit, Claude challenged me to take action (in the right direction), not just wallow in my regrets. Still, it may be true that general-purpose LLMs don't do this consistently enough.
- This! I'm tired of the constant calls to be as mediocre as we can get away with, in the name of getting things done faster and cheaper.
- Would it be feasible, with hayro-interpret and krilla, to take an existing PDF and round-trip each of the pages while wrapping the contents in marked content spans and adding tags, to remediate the accessibility of an existing PDF? Round-tripping each of the page content streams through a full-featured PDF interpreter seems cleaner than trying to edit in-place. PDFium can round-trip the content streams and add the marked content spans, but can't do the tagging. What do you think?
- Are there places where it reliably saves time, though? Unlike most software tools, LLMs are stochastic, right?
- Why do you say that? I'm in agreement with the other responses, and this thread has prompted me to start a recurring donation to the Internet Archive.
- Want to share any links, particularly about building one's own Debian image? I'm especially curious about the bootloader.
- > Are there even any social networks anymore?
Mastodon?
- > (outside of very RAM-constrained hardware, at least)
I've spent much of my career working on desktop software, especially on Windows, and especially programs that run continuously in the background. I've become convinced that it's my responsibility to treat my user's machines as RAM-constrained, and, outside of any long-running compute-heavy loops, to value RAM over CPU as long as the program has no noticeable lag. My latest desktop product was Electron-based, and I think it's pretty light as Electron apps go, but I wish I'd had the luxury of writing it all in Rust so it could be as light as possible (at least one background service is in Rust). My next planned desktop project will be in Rust.
A recent anecdote has reinforced my conviction on this. One of my employees has a PC with 16 GB of RAM, and he couldn't run a VMware VM with 4 GB of guest RAM on that machine. My old laptop, also with 16 GB of RAM, had plenty of room for two such VMs. I didn't dig into this with him, but I'm guessing that his machine is infested with crap software, much of which is probably using Electron these days, each program assuming it can use as much RAM as it wants. I want to fight that trend.
In this specific case, I don't think the rewritten version of the document is infantilizing.