RFC-0038: Separating layout from constraints

RFC-0038: Separating layout from constraints
StatusRejected
Areas
  • FIDL
Description

We propose a syntax change to convey differences between layout from constraints.
Authors
Date submitted (year-month-day)2019-03-07

Edit this RFC

Edit RFC metadata

Rejection rationale

When this proposal was drafted, and socialized, there was a strong consensus to consider syntax changes all at once, rather than one at a time (see also RFC-0039). We also wanted one person to be the syntax arbiter, rather than risk designing by committee.

Eventually, this proposal was obsoleted by RFC-0050 which met both conditions sought.

Summary

We propose a syntax change to convey differences between layout from constraints.

Motivation

Layout vs constraint

Quickly:

  • If two types have a different layout, it is not possible to soft transition from one to the other, and vice versa (not easily at least).
  • The layout describes how the bytes are laid out, vs how they are interpreted.
  • The constraint of a type is the validation step done during encoding/decoding.
  • Constraints can evolve, and as long as writers are more constrained than readers, things are compatible.

Same syntax for different things

Types which have different layout, and types which have the same layout (but different constraints) look alike.

Same layout:

  • vector<T>:6, or vector<T>:10
  • T? where T is a union, vector, or string
  • handle, handle<vmo>, or handle<channel>

Different layout:

  • array<T>:6 vs array<T>:10
  • S? where S is a struct

Design

Align on the syntax

layout:constraint

For types, i.e. anything that controls layout is before the colon, anything that controls constraint is after the colon.

Suggested changes:

    array<T>:N    becomes            array<T, N>

    S?            becomes            box<S>:nullable       (S is a struct)

    T?            becomes            T:nullable            (T is a vector, or union)

    string        is an alias for    vector<uint8>:utf8

    handle<K>     becomes            handle:K

Notes:

  • Not all constraints are meaningful for all types, for instance it is not possible to mark a struct nullable.
  • Not everything can be boxed, initially only structs can (the goal is to change the syntax, not introduce more ways to have optionality).

Ergonomics

This proposal improves ergonomics by conveying ABI implications to developers through syntax.

Documentation and Examples

At least:

Backwards Compatibility

This is not source level backwards compatible.