W3C workshop on
Web and Machine Learning

Web Platform: a 30,000 feet view

Dominique Hazael-Massieux – @dontcallmeDOM
Summer 2020

Goal

Web Platform

Web Platform specificities

Networked
User
Standards
Device Independent

Anchored in the network

Web browsers anchor their security boundaries and their trust model in the network, with a rich palette of protocols to interact seamlessly within these boundaries

USER agent

Browsers mediate user needs and wishes, including

security
privacy
accessibility
internationalization

Standards-based

The Web Platform in anchored in Royalty-Free standards defined in the open and implemented in competing products

Cross-platform

Independent of

device type
screen size
hardware capabilities
operating systems

Impact?

Intersection with Machine Learning?

Constraint: Concurrency

JavaScript is a single-threaded programming language, and in browsers, its main event loop is used fo user interactions, so running heavy computation requires care

Constraint: Same-Origin-Policy

Network sources define the security boundaries in Web browsers

Constraint: Tracking protection

Protecting users’ privacy via

Limiting uniquely identifable data (“fingerprint”)
Limiting how storage can be used by apps

Constraint: Platform neutrality

Optimize for but work without:

Different CPU set up (architecture, # of cores, power)
GPU
NPU
TPU
DSP
<insert future hardwhare here>

Web Platform Pillars

W3C can help figuring out the right trade-off between bringing more power and more features while preserving the needs and wishes from end-users

Thanks

dontcallmeDOM / dom@w3.org

Machine Learning
♥
Web Platform