Hacker News Reader

Comment by McGlockenshire | Hacker News Reader

None of the things listed as "not currently planned" are big surprises, though seeing WebRTC there is very disappointing. I've never seen "Object RTC" before, is it a competing effort, or a restandardization effort?

View on HN

"MathML" is listed as "not currently planned".

Honest question: now that mathjax has a high-quality pure-HTML/CSS output engine, does anyone still care about browser MathML support?

Object RTC is developed by a community group at the W3C. It includes people from Google, Microsoft and a number of telecom companies, amongst others http://www.w3.org/community/ortc/

so whats the difference between the two? its the first time i hear about object RTC.

Microsoft had a competing proposal to WebRTC called CU-RTC-Web which, as far as I understand it, proposed to get rid of the Session Description Protocol (which is text-based and thus difficult to parse/edit) and provide lower-level APIs instead (details here: http://www.tokbox.com/blog/what-the-cu-rtc-web-vs-webrtc-deb...).

I've just started reading up on ObjectRTC but I believe it plans to offer an API to replace SDP too and Microsoft has moved to it since it's similar to their CU-RTC-Web proposal. All in all, it looks like this will delay the adoption of WebRTC everywhere but will be for the best in terms of API and developer experience in the end.

"Lowe level API instead" suspiciously sounds like a lock in tactic for a platform specific stuff (read Windows).

No it doesnt

> which is text-based and thus difficult to parse/edit

That's not what I normally conclude when I think "text-based."

I think he means programmatically :).

It's a nicer javascript object based API for interacting with WebRTC. WebRTC by default uses an SDP offer/answer based approach.

Here is a more detailed rationale for this API: http://tools.ietf.org/id/draft-raymond-rtcweb-webrtc-js-obj-...

IMO getUserMedia is more useful than WebRTC.

Those two things are... orthogonal? WebRTC is a transport for media and data streams, which looks somewhat like SCTP-over-DTLS-over-UDP. getUserMedia is a way to get a stream which can then be sent over a transport.

Not quite - getUserMedia is one part of WebRTC. If you use the 'full' WebRTC (i.e. RTCPeerConnection), you're tied into a certain architecture, which isn't generally useful if you already have a server architecture. getUserMedia is more useful in general for developing web conferencing solutions IMO.

I think a lot of browser developers (Safari/IE) are baulking at implementing google's WebRTC. I would hope they would at least offer getUserMedia even if they don't implement the full WebRTC.

If you've looked into WebRTC you'll see it is horrendously over-engineered. I was hoping to grab the AEC code, but gave up on that idea after looking into it in more detail.

I don't see over-engineering in WebRTC; I see the same assisted-peer-to-peer routing required to support the major use-cases of traditional SIP:

1. When people are within the same firewall, they want to be able to communicate directly without routing anything but connection set-up through any servers; and

2. when people are members of the same corporation or mobile ISP, but are behind different firewalls, they want to be able to communicate by routing through the corporation/ISP's shared pool of TURN servers, instead of your service's hosted TURN servers.

Did you actually get into the internals of WebRTC on your project? That is when you notice the overengineering.

In our product, our customers generally want the session recorded, which doesn't seem possible with WebRTC. Also we need to provide group video, which peer-to-peer isn't very useful for (in terms of efficiency).

That's not to say that WebRTC isn't useful in other use-cases (such as your product). However getUserMedia is more generic because it can be used in web conferencing products like ours. That is why I say that if browser developers don't want to implement the full WebRTC, they should at least implement getUserMedia.

Ah, now I see what you mean. I agree--getUserMedia is useful on its own, with or without WebRTC.

Personally, I do hope they implement the full stack, though, because WebRTC--or ObjectRTC--is useful on its own, too (this is why I was saying they were orthogonal.) WebRTC using pure Data Channels is wonderful for enabling things like BitTorrent-like data sharing and realtime LAN gaming, whether or not there's any videoconferencing going on.

I think the proposed changes are akin to the WebSQL early on, vs. IndexedDB stuff we wound up with. I think it just comes down to people looking at WebRTC and its' short comings as well as the earlier MS/IE proposals, and coming up with something that's probably better moving forward. It sucks in terms of actually getting next gen apps out... I want my Skype killer!

Reply on news.ycombinator.com