today a sysadmin sort of person in the college of engineering that houses the school of computing where I'm a professor suggested that we might need to take down a long series of technical reports, going back to the start of the 1990s, because the PDFs are not accessible, and that it is imperative that we maximize an accessibility score that is being harmed by these pdfs.
it made me think about "So Long and Thanks for All the Fish"
@regehr Oh man. People doing harmful things to make their automated accessibility metrics go up is unfortunately a widespread thing. I get why these tools exist, they are helpful for catching mistakes—but you really can't outsource your thinking to them.
@regehr The physical items in Special Collections don't comply. We should scan it all :).
@regehr Well that's a disaster! If they absolutely need to be taken down, they need to be archived first. If you point me to where, I'll start a recursive archive for the Internet Archive. That's still not ideal, as it would become even more inaccessible, but it would preserve from destruction.
@thalia nah I'm pushing back. I have a few ideas. the suggestion to take them down is idiotic.
@regehr The two malicious complaince approaches to this impossible problem if they're only checking for PDFs are 1) Convert to postscript (trivial) 2) Find the original TeX or FrameMaker documents.
@penguin42 I'm all for malicious compliance but it's definitely not possible to find the source documents for stuff going back to the early 90s. maybe we can render papers as images.
@regehr Render to images would be trivial - but actually thinking about it, if they're from early 90s, you must have done a more recent postscript or scan to pdf - they wouldn't have been pdf that early.
@penguin42 hahaha it looks like the PS is there up through not that many years ago. so yeah maybe that's what we should do
@seachanged nice. I'll see if college IT guy can use this!
@regehr serious question, but are modern LLMs good enough to just handle this task reliably?
@krismicinski they are, but of course the university isn't giving us any help with this
Sounds like a job for an LLM.
And on searching, I find:
https://www.youtube.com/watch?v=ffe3bCcP9K8
