Re: URL.createObjectURL() for DataChannels

That's pretty cool.  Has anyone done something a little more
open-source-ish so that piranna@gmail.com and others interested can see how
it could be done?


On Tue, Jul 16, 2013 at 4:58 PM, Silvia Pfeiffer
<silviapfeiffer1@gmail.com>wrote:

> I thought that's exactly what these guys do:
> http://gigaom.com/2013/03/28/peercdn-p2p-cdn/
>
> On Wed, Jul 17, 2013 at 9:42 AM, Peter Thatcher <pthatcher@google.com>
> wrote:
> > Although it may have had some details incorrect, I think the general
> idea is
> > an interesting one:  could web page resources (such as images) be loaded
> > through a the data channel?
> >
> > I can think of a few ways that might be accomplished.  But before I think
> > too much about it:  has anyone else spent some time thinking about how it
> > could be done and under what circumstances it could be useful?
> >
> >
> >
> > On Tue, Jul 16, 2013 at 5:46 AM, Harald Alvestrand <harald@alvestrand.no
> >
> > wrote:
> >>
> >> On 07/15/2013 11:32 AM, piranna@gmail.com wrote:
> >>>
> >>> With URL.createObjectURL() you can create a temporal URL for example
> >>> from a getUserMedia() video and attach it to a <video> tag, or you can
> >>> use it with a Javascript fragment of code inside a Blob object and
> >>> give it to an inline WebWorker (also from inside a file:// scheme).
> >>
> >>
> >> Sorry, I can't parse that description.
> >> - It's "temporary", not "temporal"
> >> - There aren't Javascript fragments inside Blob objects
> >> - I don't know what you mean by "give it to an inline WebWorker".
> >>
> >>
> >>>
> >>> My idea is to be able to do the same but with WebRTC DataChannels.
> >>> This way, they could be accesed using an iframe location or similar
> >>> (also a <video> tag if required) using nested URLs this way:
> >>>
> >>> iframe.location = "http:blob:<random-blob-UUID>/index.html"
> >>
> >>
> >> This URL is malformed, unless I've forgotten everything I know about
> HTTP
> >> URLs.
> >> If you want an URL that accesses some protocol other than HTTP, please
> >> don't use HTTP URLs.
> >>
> >>
> >>>
> >>> This would allow literally surf over webpages hosted on the other peer
> >>> or whatever other thing you can do with a normal URL on the browser
> >>> without needing to reimplement them, just changing the TCP/HTTP
> >>> transport for the datachannel itself, and also maybe it would be
> >>> integrated directly on the browser (via plugin?) so it would run on
> >>> the main document setting that URL directly in the URL address bar.
> >>
> >>
> >> Unfortunately HTTP has the concept of an end-of-page (unless you use one
> >> of the infinite-document tricks that people have been using while
> waiting
> >> for webchannels to become available).
> >>
> >>>
> >>> For security, it would be required that only connect to DataChannels
> >>> whose protocol attribute is set and it's equal to the scheme we are
> >>> trying to use on it.
> >>
> >>
> >> Please try to reformulate what you want to do rather than trying to
> >> express it in terms of HTTP; HTTP just doesn't work that way.
> >>
> >>
> >>>
> >>>
> >>> --
> >>> "Si quieres viajar alrededor del mundo y ser invitado a hablar en un
> >>> monton de sitios diferentes, simplemente escribe un sistema operativo
> >>> Unix."
> >>> – Linus Tordvals, creador del sistema operativo Linux
> >>>
> >>
> >>
> >
>