- From: Simon Pieters <simonp@opera.com>
- Date: Mon, 28 Jun 2010 17:32:32 +0200
On Mon, 28 Jun 2010 16:42:45 +0200, Wellington Fernando de Macedo <wfernandom2004 at gmail.com> wrote: > > Em 28/06/2010 11:28, Simon Pieters escreveu: >>> * When at least one open or message events has been received, and >>> there are >>> close events listeners (the close event could be flagged as 'strong' >>> in this >>> case); >> >> You can get close event without an open event first. (I don't think >> Firefox gets this right currently.) > That situation only can happen (in Firefox) when the js script calls the > Close() method and the readyState == CONNECTING. > When the connection can't be established for any reason the WebSocket > object stays in the CONNECTING > readyState and doesn't dispatch any events. That behavior is because of > the section "6.1. Client-initiated closure" > of the WebSocket protocol spec, in order to prevent script to > distinguish any of the situations listed there. AFAICT, the spec requires a close event to be fired if the connection fails for any reason. It's just that the API shouldn't expose the reason for the close to Web pages. >> Also see >> http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2010-May/026400.html >> (we have implemented this policy). > Thanks! > > Wellington. > -- Simon Pieters Opera Software
Received on Monday, 28 June 2010 08:32:32 UTC