Re: Improvements suggestion for DataChannels

On 3/3/2014 9:13 AM, Harald Alvestrand wrote:
> On 03/03/2014 02:06 PM, Martin Thomson wrote:
>> On 3 March 2014 04:18, Adam Bergkvist <adam.bergkvist@ericsson.com> wrote:
>>> I still think that's strange API behavior.
>> I don't see any strangeness here.  If I have a reference, then the
>> object is not garbage collected.  If not, then not.
>>
> Do we currently close a datachannel if it's garbage collected (that is,
> if the app loses all references to it)?
> Or do we let it sit idle until the PC is garbage collected?

Absolutely it can be GC'd per the spec.

> (section 5.4 of the spec seems to allow us to garbage collect it if
> there are no handlers registered, but doesn't explicitly say that it
> closes when garbage collected).

It would be useful to add a sentence to say it's closed if it's GC'd.

> If we introduce a getter, people have a reasonable expectation that the
> channel should stick around until closed, even if they have not bothered
> to add event handlers to it yet.

You have a point, in that we usually try to avoid users knowing about GC 
- but we've already broken that for WebSockets and DataChannels.

I'll restate my comment from my response to this issue:
[ I don't have a problem with doing this, but: ]
The strongest argument against is that the functionality is already (as 
you say) implementable in userspace for those who need it without major 
trouble.

-- 
Randell Jesup -- rjesup a t mozilla d o t com

Received on Monday, 3 March 2014 14:23:42 UTC