Linux.conf.au 2020 – Wednesday – Session 2 – Unix Legacy / Social Media Research

What UNIX Cost Us by Benno Rice

Not everything is a file

Connecting to a USB device:

  • Windows – not too bad
  • Mac – a little weird
  • Linux – Lots of weird file operations. ioctl to pass data back and forth

Even worse API for creating usb_fs device. Lots of writing random data to random files.

But this is all behind a nice library?

  • Yeah but it is still a mess under the hood

Got a Byte? – Unix IO model

  • Works okay on small slow machines with simple slow interfaces
  • Doesn’t work so well with Internet, blocking
  • poll still has performance limitations
  • kevent api looked nice but Linux got epoll instead (but focuses around file descriptors)
  • But they are all still synchronous
  • Windows has Async calls

Unix is Tied to it’s history

  • Windows is newer so could learn from what came before and targetted newer hardware

C is for Colonialism

  • Farming in Europe
    • Moved to Australia, everything they new about farming doesn’t work any more.
  • PDP 11 was what Unix originally was one, simple process model.
  • Modern CPUs are not very simple
  • New CPUs lie to the OS about what the state of the machine really is (see Spectre).
  • C is not built to handle this.
  • C doesn’t handle
    • Vectorisation
    • Structure layout and padding
    • Arrays, pointers etc
  • We are not on a PDP-11 anymore
  • We have failed to evolve out CPUs and C because they are locked to each other
  • “C is not a Low Level Language” – Article

The UNIX Philosophy Problem

  • Lots of different definitions
  • Pipes seem important
  • Everything I like about using computers these days tends to be big integrated desktop tools.

Unix Suited it’s time

  • By accident it became the thing we all use
  • That time was a long time ago

How we run the community has also evolved.

Privacy is not Binary: A discussion of data systems, ethics, and human rights by Elizabeth Alpert and Amelia Radke

I was a little late to this talk so missed out the first 10-15 minutes

Social Media data reuse

  • Used by the providers
  • Governments
  • Other users
  • Malicious Users

Chucking lots of data into an “AI” is seen as yelding interesting and cool data.

Within Academia

  • Risk management. Aware of harms, mitigated, risk/reward

Is Social Media data public or private?

  • It was shared with the expectation of a certain context
  • Had to write things your friends but keep random 3rd parities in mind
  • Inferring personal information -> Dangerous
    • Especially when you are trying to infer “protected” characteristics like sexuality or religion
  • Consent? – Tricky
  • Anonymizable? – Doesn’t work

Perceptions of Risks

  • At risk groups usually given higher protection
  • Privacy is cultural concept
  • Cultural Maps

How do we do things better

  • Ethics can’t be just one person’s responsibility, it has to be in all decisions
  • Who does this belong to?
  • How do they want it to be shared?
Share