Re: Default value of "preload" attribute

On Tue, Mar 30, 2010 at 1:35 PM, Ian Hickson <> wrote:
> On Mon, 29 Mar 2010, Philip Jägenstedt wrote:
>> We haven't implemented preload yet, but I think we'll default to
>> metadata (including decoding the first frame). However, this just means
>> that we start out conservative, as soon as the the video starts playing
>> (or we have some other reason to think it will be used soon) we'll
>> download more aggressively. In other words, preload="metadata" just
>> means "conservative" and preload="auto" means "aggressive"...
> On Thu, 25 Mar 2010, Robert O'Callahan wrote:
>> Firefox defaults to "metadata". I believe that's the safe option for
>> authors who aren't sure what they're doing.
> I've changed the spec to default to metadata.
>> > > One more thing ... the resource fetch algorithm says that when
>> > > preload="metadata" UAs should stop loading after the first frame
>> > > (HAVE_CURRENT_DATA). Shouldn't that value be named "firstframe" (or
>> > > something else) then? It's pretty confusing to have
>> > > preload="metadata" not stop after the metadata has loaded.
>> >
>> > Yeah, the name is mildly misleading. I considered "firstframe" but it
>> > didn't really seem in the style of the Web platform, insofar as the
>> > Web has a style. It's not really that confusing; it's not like authors
>> > are going to say "goodness, why is it that it downloads up to the
>> > first frame instead of stopping 512 bytes earlier at the end of the
>> > metadata?". In fact you could even argue that for many purposes (e.g.
>> > filling in the poster frame if poster="" isn't provided) the first
>> > frame is a kind of metadata anyway.
>> I don't really understand why "firstframe" isn't "in the style of the
>> Web platform".
> Multiword values are generally rare, and inconsistent in their naming.
> This is highly subjective though, I agree. (Another problem with
> "firstframe" is that the media resource might not be framed based, e.g. an
> audio resource.) If there's a better term than "metadata" I'm certainly
> happy to consider it though.

Since we are only decoding the first frame to get further information,
I would almost tend to just call that "metadata", too. IMO, it's all
about setting up the decoding pipeline. Thus, I am quite happy with
the selection of the word "metadata" - it also matches with the naming
of events, I believe.

>> Loading the first frame can consume an order of magnitude more data than
>> loading the metadata. Mainly though, I'm concerned that authors will
>> want to preload up to the first frame and will think "hmm, I can preload
>> up to the metadata, but how do I preload up to the first frame?"
> If authors start thinking about this that carefully, then it's time to
> invest in porcine-safe umbrellas...

Yeah, I don't think authors will worry this much.


Received on Tuesday, 30 March 2010 03:52:15 UTC