Re: [server-timing] first run at spec draft

On 13.02.2015 20:32, Ilya Grigorik wrote:
>
> On Fri, Feb 13, 2015 at 6:38 AM, Yoav Weiss <yoav@yoav.ws 
> <mailto:yoav@yoav.ws>> wrote:
>
>     On top of what we discussed during the call, do you know of any
>     browser that intends to implement trailer support. Currently it
>     seems like there is no support for it
>     <http://www.browserscope.org/?category=network>.
>
>
> I believe most browsers "support it", in the sense that you can send a 
> trailer and it won't break anything, but we just don't do anything 
> meaningful with it - e.g. if you show a trailer we won't show it in 
> developer tools, etc. For Server-Timing I can see trailers being an 
> important and popular feature, so we'd have to get the right 
> processing logic in place.

Opera Presto supports and uses trailers. We used it for our compression 
feature (Turbo) to give information back to the client about compression 
ratio for each resource. This was before SPDY was used. In Chromium 
based Turbo all info is passed in custom frames in SPDY instead.

/Jonny

>     Also - regarding syntax, if I'm reading it correctly, why are
>     metric and description optional?
>
>
> - sometimes the name of the metric is sufficient - e.g. "edgehit" / 
> "cachehit", you could report 1/0 value but there is no reason to.
> - sometimes there is no meaningful numeric value to report - e.g. 
> "dc;atl" indicates that Atlanta DC was used.
>
> Since these are communicated via HTTP headers we want to keep things 
> terse, I don't think we should force everyone to have a mandatory 
> value and description. The only required field is the name.
>
>     And what's the use case to enable multiple descriptions on a
>     single metric?
>
>
> Hmm? That's not allowed. The format is: name=value;description, where 
> value and description are optional. That said, you *can* have multiple 
> metrics with the same name.
>
> ig
>
>