Re: Network Information API: Mozilla's proposal

On 12/29/2011 04:55 PM, Josh Soref wrote:
> This is definitely a better proposal.
>
> But it fails to consider that performance can vary by endpoint. The same can apply for metering.

To stay correct, the API can say that the returned value should be 
specific to the subdomain. Might be hard/annoying to implement though.

> Recently, I heard there was some discussion somewhere about exposing information about metering, but unfortunately no URL was provided.

Too bad, that would have been very interesting.

> For bandwidth, I believe the right API is one which provides feedback on actual network requests. I think that Progress Events [3] enable this already. There's also the WebPerf WG [4] which has a number of documents, which may address this, such as
> resource timing [5].
>
 > [3] http://www.w3.org/TR/progress-events/
 > [4] http://www.w3.org/2010/webperf/
 > [5] http://www.w3.org/TR/2011/WD-resource-timing-20110524/

I'm not sure I got what you meant. Do you think websites should use [3] 
or [5] to know about the bandwidth?

> For metering, without information from the user's ISP in the form of an API that is specific to the destination and the intended time window [6], any answer we could provide is misleading and even harmful, just as you noted the current API.
 >
> [6] http://arstechnica.com/business/news/2009/01/how-canadian-isps-throttle-the-internet.ars

[6] isn't about metering but about bandwidth limitation (or I missed 
something). If some ISP do metering for ADSL/cable/fiber connections, 
it's very likely high enough to not bother informing websites.
By metering, I mean a monthly bandwidth limit or a pay-per-use 
bandwidth, not a bandwidth speed limitation. That should be reflected by 
the bandwidth attribute.

--
Mounir

Received on Tuesday, 3 January 2012 14:43:47 UTC