1247634 - Update TCPSocket to current W3C spec

Status: RESOLVED INVALID

(Core :: DOM: Device Interfaces, defect)

Description

[Adam Roach [:abr]]

+++ This bug was initially created as a clone of Bug #1056444 +++

Several changes have been made to the API since our implementation was written. In general, these changes take two forms: conversion of EventHandlers to Promises, and the use of Streams for input and output.

Spec reference: http://www.w3.org/2012/sysapps/tcp-udp-sockets/#interface-tcpsocket

Comment 1

[Andrew Sutherland [:asuth] (he/him)]

My (concise, potentially incorrect) understanding of the situation as last discussed with :sicking and the spec author is that:
1) The spec is dead and the editor is done with it.
2) No one plans to implement it / converge to it.
3) If Mozilla made any changes to our API, we might as well just adopt the Chrome Apps socket API since you can wrap/polyfill TCP into any form you want, generally, and it's a net improvement to converge to one API.

The mooted dev-b2g and dev-webapi forums are, I believe, where the Mozilla discussions took place, if you are interested.  I can also help dig them up.

That all said, it would be totally sweet to adapt the TCP API to use the Streams API when available.

Comment 2

[Adam Roach [:abr]]

(In reply to Andrew Sutherland [:asuth] from comment #1)
> My (concise, potentially incorrect) understanding of the situation as last
> discussed with :sicking and the spec author is that:
> 1) The spec is dead and the editor is done with it.
> 2) No one plans to implement it / converge to it.
> 3) If Mozilla made any changes to our API, we might as well just adopt the
> Chrome Apps socket API since you can wrap/polyfill TCP into any form you
> want, generally, and it's a net improvement to converge to one API.

Thanks! I managed to track down some of that and deduce the rest after I filed the bug. I was thrown off by finding an editor's draft that was less than a month old, which gave the impression of active development. It appears to have been an automated build triggered by an update to the "README" in the git repo. *sigh*

Given the fact that this spec has been abandoned, I agree that making changes to our interface to conform to it doesn't make a lot of sense. We should simply document what we have, and make it available to WebExtensions.

Unless Chrome intends to expose chrome.sockets to WebExtensions, the value of implementing an act-alike API has questionable value.

I'm closing this bug as being unactionable.