Re: Ringmark, Core Mobile, and the Financial Times

Hi Jean-François,

On Jun 11, 2012, at 13:22 , <jeanfrancois.moy@orange.com> <jeanfrancois.moy@orange.com> wrote:
> When talking about speed and browsers, another factor must be taken into
> account: the hardware.

Yes, that's why I pointed out "average hardware" (though picking "reasonable" or "typical" may be better). Obviously, you can probably make just about anything run fast if you put a 1.5Ghz quad-core and some gigs of RAM into your device — if we do this, we need to make sure that it's relevant to what's out there.

> Speed is subjective and assessing that an
> implementation performs well enough on a platform without any further
> explanation might not be enough.

Yes, a lot of those measurements will be subjective. Some of them you might be able to assess (e.g. canvas fps might be measurable, at least if the API is really synchronous or close enough) but most will require a consensus — I think we can work with that though, consensus is what such groups are for (even when it's hard).

> In the same way some application users complain when an Android
> application (for example) does not perform as advertised on their
> low-end phone, we could encounter the same problem with web features (I
> am thinking particularly of CSS transitions or Canvas drawing). For that
> reason, if we happened to collect test results regarding an
> implementation speed as well as developers "bug reports", we should take
> a reference device (or hardware) for a particular platform/browser
> combination.
> 
> What do you guys think about that?

In the BrowserIQ case I would expect that we would be the ones producing the performance assessment. We'd select the hardware based on our assessment of what the target devices would be for the sort of services that we're looking at. I'm guessing for instance that when the FT made the call to go hybrid rather than pure web on Android, it wasn't because a few low end devices didn't deliver — it was likely because some mid to high end ones performed poorly (though I'll let Robert confirm here). So we should be able to drive this in a market-based fashion: look at what a project like the FT would target, and pick that as our basis.

-- 
Robin Berjon - http://berjon.com/ - @robinberjon

Received on Tuesday, 12 June 2012 10:19:01 UTC