- From: Silvia Pfeiffer <silviapfeiffer1@gmail.com>
- Date: Mon, 27 Oct 2008 23:54:53 +1100
- To: "Jack Jansen" <Jack.Jansen@cwi.nl>
- Cc: "Media Fragment" <public-media-fragment@w3.org>
Hi Jack, all, On Mon, Oct 27, 2008 at 11:31 PM, Jack Jansen <Jack.Jansen@cwi.nl> wrote: > On 27 okt 2008, at 12:11, Silvia Pfeiffer wrote: > >> Wow, that's totally awesome. I am inspired to write the CMML/Annodex >> section. :) > > You know, that would be totally awesome too:-) OK, done. I took the same headings that you used for SMIL and described how we do it with Ogg. We have a lot more different components to the technology (CMML, Ogg skeleton, temporal URIs, ROE), but I hope I haven't confused anybody. BTW: Davy - I'd be curious if your meta-specification format of the structure of audio & video could be mapped into ROE somehow... >> BTW: I like the quadruple way of specifying a spatial fragment (area) >> in SMIL: x-offset, y-offset, width, height. We should consider using >> that for spatial media fragment URIs. > > There's a problem with 4-tuples for rectangles (that I've already touched > upon in the piece of text) and that is that sometimes it's intended to be > x,y,w,h and sometimes it's l,t,r,b (or x1,y1,x2,y2 which is usually the > same). > > In SYMM this bothered us to no end, because we wanted to be compatible with > the spec from which we lifted the feature, but this meant we couldn't be > internally consistent anymore:-( The SMIL native method is to spell things > out: region boundaries can be specified with attributes > top,left,right,bottom,widht,height, with all values defaulting to "auto". So > as long as you don't overspecify any dimension you're fine. > > In hindsight, I think it might have been better not to use any four-tuples > but in stead spell things out (so, in stead of panZoom="25%,25%,50%,50%" use > clipLeft="25%" clipTop="25%" clipWidth="50%" clipHeight="50%"). > > Also note that <area> *must* use ltrb-style to be consistent: if the shape > is a polygon you must specify x0,y0,x1,y1,x2,y2, ... anyway, so if you don't > do a point pair for a rectangle (but in stead a point, size pair) things > become messy. I see where you're coming from. I'd still prefer having just one parameter and however many values we need after that to keep it compact in a URI. But that's just me and up for discussion. :-) Cheers, Silvia.
Received on Monday, 27 October 2008 12:55:30 UTC