The difference between audio and video is that i think your ears will put up with lower quality than your eyes are willing to.
Actually, the converse is true. Research shows people will quit watching something with bad audio and good video way before they quit something with bad video and good audio. Hence YouTube.
By example, a dropped frame here and there isn't that horrible, but if you lose a 1/24th of a second of audio every few seconds, you'll bail out pretty quickly.
It's just that the bitrate required to saturate our requirements for audio quality are trivially achievable with modern compression and bitrates. WMA 10 Pro and HE AAC v2 @ 64 Kbps are pretty much transparent for most people with most content. the 256 Kbps AAC-LC used in iTunes is well past the "entertainment-grade with good headphones" point - 128 Kbps would be fine most of the time, and 192 Kbps nearly all the time.
The last-mile physical wiring hasn't improved much over the last 10 years, and I doubt it will over the next 10. Fiber to the home ain't coming soon.
Eh? 10 years ago, I had the best DSL money could buy - 640 Kbps for $120/month. Today I have the best DSL money can buy - 20 Mbps for $60/month. I think a 30x improvement in a decade is pretty darn material:)! Although the copper to my house isn't different, Qwest has brought in fibre to the node, so that there's a much shorter loop between my house and my internet access.
When we move over the summer, I'll have 50 Mbps cable.
20 Mbps is PLENTY to delivery perfect 1080p quality. Really, 10-12 Mbps is enough for most movie content, and it'll be 8-10 Mbps next year due to further codec improvements.
Here's a sample I did a couple months ago with some widescreen movie standard test content at 6 Mbps average 8 Mbps peak, including 5.1 audio. Other than the opening "confetti credits" shot, which is purposely designed as a codec-buster, the quality is quite good throughout.
Given the intersecting curves of improving bandwidth and improving codecs, 1080p delivery to more than half of USA households should be possible in 2010.
There's no 'Display PDF' on OS X, you've probably mistaken it with Display Postscript from Classic Mac OS era. What are you saying might have been true for Display Postscript, but OS X's Quartz 2D is something completely different.
Display PostScript was from OpenStep. Display PDF is a nickname for the imaging model used in Quartz on Mac OS X.
Apple's Mac OS X operating system uses a central window server (created entirely by Apple) that caches window graphics as bitmaps, instead of storing and executing PostScript code. A graphics library called Quartz 2D provides PostScript-style imaging using the PDF graphics primitives (a superset, plus tweaks, of the PostScript model), but this is used by application frameworks--there is no PostScript or PDF present in the Mac OS X window server. Apple chose to use this model for a variety of reasons, including the avoidance of high Adobe-imposed licensing fees for DPS, and more efficient support of legacy Carbon and Classic code; QuickDraw-based applications use bitmapped drawing exclusively. Adobe's copyright stipulations for the PDF standard are much less restrictive, granting conditional copyright permission to anyone to use the format in software applications, free of charge.
They don't need to. It follows that if the movie is too slow, try the smaller one. Alternatively, they can try the HD version and see if it works for them. There's more steps, but it's not a burden by any stretch of the imagination.
Perhaps we have different operating definitions of burden here. Having a user try one thing, get a bad expereince with potentially unclear feedback, then try another thing? Not optimal, and worth fixing. Plus users are as likely to click on the loweset quality experience "just because it'll work" even if a higher one would have worked. This is a bigger deal with longer-form content and real-time playback, as changing networking conditions, PC load, etcetera can mean what was optimal when something was started does't remain optimal throughout the whole experience.
Why put the burden of that on the user; the application can have a much better sense of what it can do that the user.
I honestly think you're trying to solve a problem that doesn't exist. Balancing the movie quality to reach the highest number of users appears to be the best option here. iTunes users seem more than happy enough with the service. (I've certainly never had any problems with movie quality.) Do those users have a burning need for downloadable movies in 1080p? Hell, no. And you're not going to provide it to them anyway. (See below)
Great if iTunes has worked for you. But that's an explicitly download experience. In particular for rental content, forcing a long download for content that's going to expire anyway seems like a waste of user time. Much better to let them watch immediately if they want to.
It's interesting how you understand this without actually understanding it. By your own admission, a user is more likely to try and find the video they want on Youtube long before they become willing to download a Silverlight plugin. Only after they have exhausted all accessibility options for the content they want will they accept the quality benefits. In other words, you've just argued against your own product.
We'll obviously need to have Silverlight to be broadly installed, and very easy to istall for those who don't have it. That's something we're engaged with on a variety of fronts, and orthogonal to the issue of the accessibility of the experience once Silverlight is installed. Smooth Stremaing is still in alpha anyway; Silverlight will have that much higher market share by the time it's launched.
Actually, you're going to provide them with a substandard experience. You're automatically going to degrade their experience based on factors that the user doesn't understand. The user can do nothing to change the settings and find for himself why his experience is worse than his neighbor's experience. It just is. He's painfully upgrade all his hardware and STILL have a worse experience, never knowing that his neighbor's bandwidth is better.
Silverlight is turing complete; there's all kinds of ways to offer users information on what's going on and what could be done to improve it, which publishers can supply as appropriate. Certainly, they'll get a better experience streaming Smooth Stremaing than they've ever had before. Everything that'd make Smooth Streaming downshift would have just killed streaming dead anyway, so having video and audio playing back without a pause will be a big upgrade for them, even if they lose some visual detail.
What kind of demo site? As in, "Now that you have Silverlight installed try this" type of site or "look what we can do with this technology you should really get" type of site? Because it fails miserably at the latter. If I can't see it, I'm not sold on the technology. If it's the former, then it's on you to sell it to Slashdot so that we're convinced to download the plugin to try it.
You're in the small minority of people who go to the site and neither have Silverlight nor install
Cross-platform....... Using a blob of codecs that are almost certainly not going to be updated or even there in a few years' time.
Moonlight was prototyped with ffmpeg. Moonlight proper is GPL; you can plug whatever media pipeline you want in there.
We're offering the binary codec download so that Microsoft can take care of the patent licensing fees for the codecs. That addresses a big concern some have about media playback on Linux.
We're planning on doing an update to the codec module including the Silverlight 3 codecs (adding H.264 and AAC-LC).
Note that the Moonlight 1.0 codec module is actually the Silverlight 2 codecs, inclduing WMA 10 Pro and a bunch of optimiation work. Contrary to your intuition, the one part Microsoft actualy delivers is the full Silverlight 2 implementation.
Interesting concept, but I just can't imagine internet connections (even on broadband) being predictable enough for this to work. My connections occasionally hiccup with a 10-20 pause; not often enough to be really annoying, but what is such a predictor going to do with that?
The nice thing is that all the stream switching heuristics live in the managed code. So tweaking can be done continuously and customized per site if appropriate. Since all the logic gets loaded with the web page, it can be updated daily if desired.
In your case, the current heuristics tries to buffer up to 20 seconds of future video, so if it only goes down for 20 seconds, you should be okay, although it'll probably throttle down your bitrate after it does't get any chunks for a while.
But that 20 seconds is just a constant in the managed code. If there's a lot of users with connections like yours, it'd be raise that to 30 seconds, or note the pauses and then raise it to 40 seconds, or whatever. The player can report back QoS data to the content publisher so they can see if, when, and where users are having any issues.
The full C# source code for the client module is available (and bundled with Expression Encoder 2 SP1).
Complete nonsense. The idea of "SD" vs. "HD" has been so ingrained into our culture at this point, that it's quite easy for users to figure it out. Take this movie website as an example:
But does the median user really know which clip their machine is capable of?
Note that Apple's trailers default to a small embedded version, presumably because their user research shows that is more accessible. The high quality download versions require a few extra clicks, and an impossing array of choices. It's a bad user experience, and it's now possible to provide a better one.
Moreso, progressive download movie trailers are great for 90-150 seconds of high interest content. Lots of trailers are worth waiting 10 min for, which can work if the data rate is 5x the connection speed. But for a two hour movie, does anyone want to wait 10 hours? Smooth Streaming's sweet spot is ffor longer form content where the user wants instant-on, low-delay scrubbing, no "buffering" messages, and at the best quality they can get within those constraints.
Again, think "PVR in the cloud".
You're completely missing the point. Multibitrate does not matter. Consider how many people link off to the Youtube versions of the Star Trek trailers! Those are of terrible quality. Yet the convenience and real-world benefits are more important to them than HD resolutions.
Sure it does. In general, content > accessibility > quality. Most people will watch content they like over content they don't like, and will put up with accessibility barriers before they'll watch stuff they don't like. Most people will put up with quality issues in order to have a much better accessibility to the content, but will chose the accessible version over the non-accessible version if both are available. Assuming the same content is availabe and accessable in a high quality and a low quality form, they'll pick the higher quality version.
Also, poor video and audio quality is very dependent on the content and the experience the user is trying to have. Sure, the PS3 has YouTube, but watching it on a HD display at leaning back on a couch at 10' is nearly unbearable, even though the same clip would be fine in a small window on a PC with other stuff going on. With Smooth Streaming, we want to really enable to full consumer HD experience AND the streaming model at the same time.
HD will catch on when the hardware gets here. And the reason why it will catch on when the hardware gets here is because that is when the best experience can be offered. It's not about HD vs. non-HD. That's a red herring. It's about providing a better service overall. HD video is a bonus and nothing more.
The hardware and broadband is here, it's just unevenly distributed. The goal of the technology is to provide every user with the best experience they can get at that moment, without having to underserve the high end or block out the low end.
You aren't listening. I don't care about the install option. I don't want to install Silverlight. It's the job of the website to convince me that "Yes you do!" The website does nothing to convince me. It merely gives me an ultimatum: We won't show you what we're about until you install this plugin. So install it or leave.
This is a technology demo site, not a consumer site. Consumer sites could deliver this in a wide variety of different ways, including offering a fallback experience if appropriate.
You meant that they could do a better job of forcing you to use Microsoft technology that's only supported by Microsoft and 'strongly encourages' users to use Microsoft Windows, all while giving them a warm/glowy feeling about Microsoft's super-duper attempts at supporting Macs in addition to the many multiple versions of Windows. (The latter being the classic Microsoft definition of "cross-platform".) That's what you really meant, wasn't it? That's the definitio
You're basically baking in the limitations of existing technology in your assumptions there.
The whole point of this is that you could deliver high-value content like movies via a streaming model without having to wait for a download. Sure, there's the airplane ride case where you need a download, but there's lots of consumers with an always on internet connection, and Smooth Streaming could offer very much a "PVR in the cloud" experience where they can watch anything immediately.
And while different content has different quality requirements, I'd say we're a long way from saturating that even for news and soap operas. Even 480i has 720x480p60 or 720x576p50 visual information, and at best users are probably seeing 640x360p25 or p30 today with streaming. So that's 2/3rd of the visual information being left on the floor today compared to what a CRT TV could deliver decades ago. So much online video today is PAINFUL to watch; we should be delivering as much visual information as there is in the source that we can.
The same is true of live - "you take what you're given"? What not give people as much as they can take? Live events can be HD as much as on-demand content, and should for appropriate content.
And if you've got a web cache that falls down with HUNDREDS of cached files in it, you're probably running Netscape Navigator 2.0 on a Quadra 700. Try looking up how many cached files you have right now; you'll be suprised.
Also, the value isn't so much in using the local browser cache, but proxy caches at ISPs and firewalls, which have a ton of storage capacity.
No, this requires the heuristics and MediaStreamSource API, both of which require managed code. This would be a Moonlight 2.0 feautre.
And do you need to have a windows streaming server to use it?
We've implemented the server side as an IIS7 module running on Windows Server 2008.
I would also love to see a hi-res (720p or higher) stream with 4-5 mbit max bandwidth, just to see how well it handle that content (both player and streamer).
The highest bitrates at SmoothHD.com are 3.5 Mbps 1280x720. You may need to maximize your browser window or go full-screen to see it. The maximum stream is determined by CPU, network speed, and screen size, so you'll want your window big enough for the media rectangle to be 1280x720.
I'm sure MS has a patent on this. So why would I want to use something that will *always* work better on Windows and put Linux developers in the cross-hairs?
The client-side logic for Smooth Streaming is all in managed code, using published APIs. Moonlight should be able to support it fine. IF you're worried about this, jump on the Moonlight 2.0 effort.
If the system doesn't fall back to something actually standards-compliant without me having to do anything, then it is a gigantic fuckup. I don't want to do it Microsoft's way, I want to do it the right way, the standard way, the correct way.
What will HTML5 fall back to when a non-HTML5 compatible browser is being used?
Fallbacks are an orthogonal and unavoidable issue. Really, the only media format that'll really play from a browser on 99% of PCs today is a http://www.foo.mpg/ link to a MPEG-1 file. It's great to fantasize about "one format to rule them all" but even when that's introduced in a product, it won't become universal until there aren't any systems left that can't play it. And the best way to embed video will be well beyond HTML5 by the time HTML5 may be universal. Heck, if the baseline is just Theora + Vorbis, it's already 5-8 years behind the state of the art. And we'll proabably have H.265 in 4-5 years...
Do what exactly? Linking off to a site that requires Silverlight with no explanation doesn't seem like a very good argument when you're posting to a forum that doesn't want to install your plugin. (Or more to the point, many of them can't install Silverlight 2.0.)
The second link is about how the stuff in the first is authored, and doesn't require Silverlight. Sorry if that wasn't clear.
Your second link talks about multi-bitrate encoding. Which strikes me as (like the entire Silverlight platform) a solution looking for a problem. Despite the fact that Microsoft has had the technology deployed for years as part of WiMP, the market hasn't bought into it. It's just as easy (and probably less confusing) to simply provide different sizes. 95%+ of current streaming videos don't even have to worry about that.
The challenge with offering multiple sizes is that in forces the user to know what their system and connection can play, and it really only works with progressive download models, not real instant-on, easy random access long-form "streaming." That's fine for some audiences, but not for the mass market. Multibitrate done right means nearly instant startup and gapless playback, dynamically adjusted to what the user's machine can play back. It's a very different use model than YouTube.
Perhaps Silverlight will be better positioned when HD streaming becomes the norm. More likely however, is that HD will be the norm when the majority of hardware on the market is both capable of HD streaming and integrated into the standard home in a way that would make HD streaming a superior enough experience for consumers to want to use it.
Ah, that's the point! Smooth Streamings gets us out of having to wait for everyoen to be able to do HD to use it for mass audience content. If only the top 40% of users can get full 720p, the top 40 % of users get full 720p. And users who have less get the best experience their hardware and network is capable off. We don't have to sweat the lowest common denominator.
As an aside, why is it that every Silverlight website stops you cold? There's not even a description of what it is you're missing and/or why you should install the plugin. It's simply "install this or go away". So I go away. No skin off my nose.
There's a lot of flexibility in how a site can present the install option. For example, NBCOlympics.com offered a fallback to an IE embedded WMP ActiveX component. I agree that more sites could do a nicer job of it, and we're talking to them about improving that experience.
And I don't see anything in Silverlight that isn't similarly addressed by HTML5. Ergo, HTML5 is superior for its standardization, true cross-platform support, and competing implementations that can meet the needs of many different ideals.
As I said elsewhere, HTML5 has no way to do anything like this:
As for cross-platform, Moonlight 2.0 should be able to run SmoothHD just fine, and more importantly a whole lot of content published using that platform.
Which is good, as VLC and thus presumably ffmpeg hasn't been able to play VC-1 files with B-frames for years, which is pretty much the default these days.
I encoded the "Big Buck Bunny" clip up there:). It's still in pre-alpha, but you should be able to get the idea
This uses a new API called MediaStreamSource, which enables file parsers and protocols to be built in managed code, and then hand off the video and audio bitstreams to Silverlight's built int decoders.
In the case of Smooth Streaming, every two seconds of the video is a seperate http request, and each of those chunks is available in six different data rates. Managed code heuristics running inside of Silverlight dynamically pick the right bitrate for the next chunk based on available CPU power, network speed, and window size (no reason to download 720p if the brower window is shrunk down in a corner of the stream).
And because this is based around small http requests, chunks get proxy cached, so 100 people watching the same video behind the same firewall would only need to get a single copy, providing much better scalability than traditional unicast streaming.
Anyway, this is something that Flash certainly can't do, and I haven't seen any hint of HTML5 being able to do. Pulling it all together requires some pretty specific characteristics of the video decoder (the ability to switch resolutions with a new sequence header without any pause), an API like MediaStreamSource, and having a performant enough runtime to be able to run all the heuristics and parsing without using much CPU.
But when you look at it any more than superficially, every dollar the government spends is taken at gunpoint (perhaps metaphorically) from someone.
The difference between being threatened with execution and having ones wages garnished is rather more than metaphorical! A just society can and should be financed without shooting anyone's grandmother.
Even incarceration is reserved for those who file fraudulent taxes, not those who fail to pay for one reason or another.
No it doesn't; it increases the profit margin of the people providing the goods. Why on Earth would they lower prices when they're making good money? Demand is not perfectly elastic in the real world (especially for necessities like food, housing, health care, etc.), so lowering the price won't necessarily make a seller more money by increasing the amount demanded. Therefore, it's in their best interests to just pocket the money, and that's exactly what they do.
Interesting theory.
Fortunately, we have rather a lot of empirical evidence on this topic! For your testable hypotheses:
Hypothesis 1: The price of imported goods imported from a low cost country will not go down. Results: They in fact went down like crazy. Have you seen what a kid's $20 toy can do these days? Shopped for discount shoes? You can buy decent quality stuff for much cheaper in inflation adjusted dollars than you used to be able to.
Hypothesis 2: Earnings of companies who largely import good will go up in proportion to the savings of manufacture in low-wage countries, less the copy of transport. Results: Nope. US corporations have pretty consistant average profits. This is predicted by economic theory. if a 5% return on capital is considered "good enough" by the market, and a company figures out how to get a 50% return on capital one year, every company in the same industry would try to copy the same technique and would be very happy to get only 20% (4x what they had before) in order to get market share, until return on capital eventually gets back to 5%. The way to be hugely profitable is to do something no one else can do, not to do what everyone else is doing a little cheaper.
Couple that with the enormous capacity increases you're seeing in regular MP3 players, and there's not much reason to go for Ogg anymore. Encode your stuff as 192kbps MP2, and it's future proof, playable on free players, playable on virtually every portable player, and higher quality than MP3 at the same bitrate. Go figure.
Er, NO. While perhaps MP3 @ 384 Kbps can outperform MP3 @ 320 Kbps (never tested that myself), MP3 mainfestly outperforms MP2 at the data rates where you can hear a difference. The old rule of thumb is that MP3 offers about 50% better efficiency than MP2, so that 192 Kbps MP2 will sound roughly like 128 Kbps MP3. Which I don't find high enough for music listening, particularly in CBR.
I don't know if I should laugh or cry. On the one hand, $100,000 is serious money. On the other hand, it barely pays for a good developer for one year.
A good, experienced encoder optimization engineer would cost a lot more than that as a contractor or as an employee ($100K salary, maybe, but a whole lot more with benefits, overhead, and decent quality gear to evaluate video quality with).
The challenge for Theora is to get competitive "enough" ("enough" being specific to use ) versus H.264 and VC-1 not as they are today, but as they'll be in the timeframe of those improvements. There's a whole lot more then $100K being spent by a variety of companies and groups to squeeze improvements out of the commonly used standard codecs as well.
Theora is a weird enough of a design I don't have a clear intuition of how good it could be. Lacking bidirectional prediction (B-frames) is a pretty big limitation out of the gate. But there's no doubt lots of room for it to be improved compared to where it is now.
Yes. Since a game is almost the canonical exclusive mode task, the OS should add as little overhead as possible, with as many cycles as possible going the the CPU and GPU.
If one version of Windows had a lot higher performance than another for gaming, that'd be more suggestive of a bug or driver issues in the slower version.
It's an easy thing to fix - require that H1B visa holders receive the same pay and benefits for their work as the rest of the workforce. If companies really have problems finding citizens to fill jobs, and aren't just trolling for lower paid wage slaves, then it ought not to be a problem, right?
Working at Microsoft, this is how it's always seemed to me. I've worked with some people for serveral years before finding out that their immigration status isn't what I'd assumed. Beyond some people having to go wait in line at the DHS for paperwork ocassionaly, I haven't noticed any practical distinction based on status. And I'd consider it rude and inappropraite if I did.
I certainly agree with your proposal, but that's essentialy what's happening already from my experience at Microsoft from what I've seen.
And there's absolutely technical sub-disciplines where there simply aren't enough US graduates to fill open positions. From my own world, signal processing is a big one. We see a roughly even mix between North America, greater China, ex-USSR, and ex-Raj origins for people who work on encoder and decoder stuff. Lots (most?) of the people who have been around for long enough have become citizens.
If anything the right solution would be to just not have a cap for skilled workers so that a seperate H1B program isn't needed. This is how Canada does it (you've got the right skills, etcetera, and you can come), and they're a better country for it.
Why are we still talking about this? Gutmann had never run Vista when he wrote the article, and stopped updating it once Vista was in the market and actually testing it was possible.
He makes a ton of testable hypothoses in the article. And I don't see a one that hasn't been disproven.
It's simply not an article about Vista. It's an article describing his fantasy of how bad Vista could have been.
Slashdot Mirror
The difference between audio and video is that i think your ears will put up with lower quality than your eyes are willing to.
Actually, the converse is true. Research shows people will quit watching something with bad audio and good video way before they quit something with bad video and good audio. Hence YouTube.
By example, a dropped frame here and there isn't that horrible, but if you lose a 1/24th of a second of audio every few seconds, you'll bail out pretty quickly.
It's just that the bitrate required to saturate our requirements for audio quality are trivially achievable with modern compression and bitrates. WMA 10 Pro and HE AAC v2 @ 64 Kbps are pretty much transparent for most people with most content. the 256 Kbps AAC-LC used in iTunes is well past the "entertainment-grade with good headphones" point - 128 Kbps would be fine most of the time, and 192 Kbps nearly all the time.
The last-mile physical wiring hasn't improved much over the last 10 years, and I doubt it will over the next 10. Fiber to the home ain't coming soon.
Eh? 10 years ago, I had the best DSL money could buy - 640 Kbps for $120/month. Today I have the best DSL money can buy - 20 Mbps for $60/month. I think a 30x improvement in a decade is pretty darn material :)! Although the copper to my house isn't different, Qwest has brought in fibre to the node, so that there's a much shorter loop between my house and my internet access.
When we move over the summer, I'll have 50 Mbps cable.
20 Mbps is PLENTY to delivery perfect 1080p quality. Really, 10-12 Mbps is enough for most movie content, and it'll be 8-10 Mbps next year due to further codec improvements.
Here's a sample I did a couple months ago with some widescreen movie standard test content at 6 Mbps average 8 Mbps peak, including 5.1 audio. Other than the opening "confetti credits" shot, which is purposely designed as a codec-buster, the quality is quite good throughout.
http://silverlight.services.live.com/31260/StEM%206-8%20Mbps%201280x800p24/video.wmv
Given the intersecting curves of improving bandwidth and improving codecs, 1080p delivery to more than half of USA households should be possible in 2010.
There's no 'Display PDF' on OS X, you've probably mistaken it with Display Postscript from Classic Mac OS era. What are you saying might have been true for Display Postscript, but OS X's Quartz 2D is something completely different.
Display PostScript was from OpenStep. Display PDF is a nickname for the imaging model used in Quartz on Mac OS X.
http://en.wikipedia.org/wiki/Display_postscript#Modern_Derivatives
Apple's Mac OS X operating system uses a central window server (created entirely by Apple) that caches window graphics as bitmaps, instead of storing and executing PostScript code. A graphics library called Quartz 2D provides PostScript-style imaging using the PDF graphics primitives (a superset, plus tweaks, of the PostScript model), but this is used by application frameworks--there is no PostScript or PDF present in the Mac OS X window server. Apple chose to use this model for a variety of reasons, including the avoidance of high Adobe-imposed licensing fees for DPS, and more efficient support of legacy Carbon and Classic code; QuickDraw-based applications use bitmapped drawing exclusively. Adobe's copyright stipulations for the PDF standard are much less restrictive, granting conditional copyright permission to anyone to use the format in software applications, free of charge.
They don't need to. It follows that if the movie is too slow, try the smaller one. Alternatively, they can try the HD version and see if it works for them. There's more steps, but it's not a burden by any stretch of the imagination.
Perhaps we have different operating definitions of burden here. Having a user try one thing, get a bad expereince with potentially unclear feedback, then try another thing? Not optimal, and worth fixing. Plus users are as likely to click on the loweset quality experience "just because it'll work" even if a higher one would have worked. This is a bigger deal with longer-form content and real-time playback, as changing networking conditions, PC load, etcetera can mean what was optimal when something was started does't remain optimal throughout the whole experience.
Why put the burden of that on the user; the application can have a much better sense of what it can do that the user.
I honestly think you're trying to solve a problem that doesn't exist. Balancing the movie quality to reach the highest number of users appears to be the best option here. iTunes users seem more than happy enough with the service. (I've certainly never had any problems with movie quality.) Do those users have a burning need for downloadable movies in 1080p? Hell, no. And you're not going to provide it to them anyway. (See below)
Great if iTunes has worked for you. But that's an explicitly download experience. In particular for rental content, forcing a long download for content that's going to expire anyway seems like a waste of user time. Much better to let them watch immediately if they want to.
It's interesting how you understand this without actually understanding it. By your own admission, a user is more likely to try and find the video they want on Youtube long before they become willing to download a Silverlight plugin. Only after they have exhausted all accessibility options for the content they want will they accept the quality benefits. In other words, you've just argued against your own product.
We'll obviously need to have Silverlight to be broadly installed, and very easy to istall for those who don't have it. That's something we're engaged with on a variety of fronts, and orthogonal to the issue of the accessibility of the experience once Silverlight is installed. Smooth Stremaing is still in alpha anyway; Silverlight will have that much higher market share by the time it's launched.
Actually, you're going to provide them with a substandard experience. You're automatically going to degrade their experience based on factors that the user doesn't understand. The user can do nothing to change the settings and find for himself why his experience is worse than his neighbor's experience. It just is. He's painfully upgrade all his hardware and STILL have a worse experience, never knowing that his neighbor's bandwidth is better.
Silverlight is turing complete; there's all kinds of ways to offer users information on what's going on and what could be done to improve it, which publishers can supply as appropriate. Certainly, they'll get a better experience streaming Smooth Stremaing than they've ever had before. Everything that'd make Smooth Streaming downshift would have just killed streaming dead anyway, so having video and audio playing back without a pause will be a big upgrade for them, even if they lose some visual detail.
What kind of demo site? As in, "Now that you have Silverlight installed try this" type of site or "look what we can do with this technology you should really get" type of site? Because it fails miserably at the latter. If I can't see it, I'm not sold on the technology. If it's the former, then it's on you to sell it to Slashdot so that we're convinced to download the plugin to try it.
You're in the small minority of people who go to the site and neither have Silverlight nor install
Cross-platform....... Using a blob of codecs that are almost certainly not going to be updated or even there in a few years' time.
Moonlight was prototyped with ffmpeg. Moonlight proper is GPL; you can plug whatever media pipeline you want in there.
We're offering the binary codec download so that Microsoft can take care of the patent licensing fees for the codecs. That addresses a big concern some have about media playback on Linux.
We're planning on doing an update to the codec module including the Silverlight 3 codecs (adding H.264 and AAC-LC).
Note that the Moonlight 1.0 codec module is actually the Silverlight 2 codecs, inclduing WMA 10 Pro and a bunch of optimiation work. Contrary to your intuition, the one part Microsoft actualy delivers is the full Silverlight 2 implementation.
Interesting concept, but I just can't imagine internet connections (even on broadband) being predictable enough for this to work. My connections occasionally hiccup with a 10-20 pause; not often enough to be really annoying, but what is such a predictor going to do with that?
The nice thing is that all the stream switching heuristics live in the managed code. So tweaking can be done continuously and customized per site if appropriate. Since all the logic gets loaded with the web page, it can be updated daily if desired.
In your case, the current heuristics tries to buffer up to 20 seconds of future video, so if it only goes down for 20 seconds, you should be okay, although it'll probably throttle down your bitrate after it does't get any chunks for a while.
But that 20 seconds is just a constant in the managed code. If there's a lot of users with connections like yours, it'd be raise that to 30 seconds, or note the pauses and then raise it to 40 seconds, or whatever. The player can report back QoS data to the content publisher so they can see if, when, and where users are having any issues.
The full C# source code for the client module is available (and bundled with Expression Encoder 2 SP1).
How is VS terrible for this?
Anyway, if you're an Eclipse fan, have you checked out the Eclipse Tools for Microsoft Silverlight?
http://www.eclipse4sl.org/
Complete nonsense. The idea of "SD" vs. "HD" has been so ingrained into our culture at this point, that it's quite easy for users to figure it out. Take this movie website as an example:
But does the median user really know which clip their machine is capable of?
Note that Apple's trailers default to a small embedded version, presumably because their user research shows that is more accessible. The high quality download versions require a few extra clicks, and an impossing array of choices. It's a bad user experience, and it's now possible to provide a better one.
Moreso, progressive download movie trailers are great for 90-150 seconds of high interest content. Lots of trailers are worth waiting 10 min for, which can work if the data rate is 5x the connection speed. But for a two hour movie, does anyone want to wait 10 hours? Smooth Streaming's sweet spot is ffor longer form content where the user wants instant-on, low-delay scrubbing, no "buffering" messages, and at the best quality they can get within those constraints.
Again, think "PVR in the cloud".
You're completely missing the point. Multibitrate does not matter. Consider how many people link off to the Youtube versions of the Star Trek trailers! Those are of terrible quality. Yet the convenience and real-world benefits are more important to them than HD resolutions.
Sure it does. In general, content > accessibility > quality. Most people will watch content they like over content they don't like, and will put up with accessibility barriers before they'll watch stuff they don't like. Most people will put up with quality issues in order to have a much better accessibility to the content, but will chose the accessible version over the non-accessible version if both are available. Assuming the same content is availabe and accessable in a high quality and a low quality form, they'll pick the higher quality version.
Also, poor video and audio quality is very dependent on the content and the experience the user is trying to have. Sure, the PS3 has YouTube, but watching it on a HD display at leaning back on a couch at 10' is nearly unbearable, even though the same clip would be fine in a small window on a PC with other stuff going on. With Smooth Streaming, we want to really enable to full consumer HD experience AND the streaming model at the same time.
HD will catch on when the hardware gets here. And the reason why it will catch on when the hardware gets here is because that is when the best experience can be offered. It's not about HD vs. non-HD. That's a red herring. It's about providing a better service overall. HD video is a bonus and nothing more.
The hardware and broadband is here, it's just unevenly distributed. The goal of the technology is to provide every user with the best experience they can get at that moment, without having to underserve the high end or block out the low end.
You aren't listening. I don't care about the install option. I don't want to install Silverlight. It's the job of the website to convince me that "Yes you do!" The website does nothing to convince me. It merely gives me an ultimatum: We won't show you what we're about until you install this plugin. So install it or leave.
This is a technology demo site, not a consumer site. Consumer sites could deliver this in a wide variety of different ways, including offering a fallback experience if appropriate.
You meant that they could do a better job of forcing you to use Microsoft technology that's only supported by Microsoft and 'strongly encourages' users to use Microsoft Windows, all while giving them a warm/glowy feeling about Microsoft's super-duper attempts at supporting Macs in addition to the many multiple versions of Windows. (The latter being the classic Microsoft definition of "cross-platform".) That's what you really meant, wasn't it?
That's the definitio
You're basically baking in the limitations of existing technology in your assumptions there.
The whole point of this is that you could deliver high-value content like movies via a streaming model without having to wait for a download. Sure, there's the airplane ride case where you need a download, but there's lots of consumers with an always on internet connection, and Smooth Streaming could offer very much a "PVR in the cloud" experience where they can watch anything immediately.
And while different content has different quality requirements, I'd say we're a long way from saturating that even for news and soap operas. Even 480i has 720x480p60 or 720x576p50 visual information, and at best users are probably seeing 640x360p25 or p30 today with streaming. So that's 2/3rd of the visual information being left on the floor today compared to what a CRT TV could deliver decades ago. So much online video today is PAINFUL to watch; we should be delivering as much visual information as there is in the source that we can.
The same is true of live - "you take what you're given"? What not give people as much as they can take? Live events can be HD as much as on-demand content, and should for appropriate content.
And if you've got a web cache that falls down with HUNDREDS of cached files in it, you're probably running Netscape Navigator 2.0 on a Quadra 700. Try looking up how many cached files you have right now; you'll be suprised.
Also, the value isn't so much in using the local browser cache, but proxy caches at ISPs and firewalls, which have a ton of storage capacity.
Does this support silverlight 1 (aka moonlight)?
No, this requires the heuristics and MediaStreamSource API, both of which require managed code. This would be a Moonlight 2.0 feautre.
And do you need to have a windows streaming server to use it?
We've implemented the server side as an IIS7 module running on Windows Server 2008.
I would also love to see a hi-res (720p or higher) stream with 4-5 mbit max bandwidth, just to see how well it handle that content (both player and streamer).
The highest bitrates at SmoothHD.com are 3.5 Mbps 1280x720. You may need to maximize your browser window or go full-screen to see it. The maximum stream is determined by CPU, network speed, and screen size, so you'll want your window big enough for the media rectangle to be 1280x720.
I'm sure MS has a patent on this. So why would I want to use something that will *always* work better on Windows and put Linux developers in the cross-hairs?
The client-side logic for Smooth Streaming is all in managed code, using published APIs. Moonlight should be able to support it fine. IF you're worried about this, jump on the Moonlight 2.0 effort.
http://on10.net/blogs/benwagg/Moonlight-10-is-released/
Also, there is a covenant not to sue if that's your concern.
http://www.microsoft.com/interop/msnovellcollab/moonlight.mspx
If the system doesn't fall back to something actually standards-compliant without me having to do anything, then it is a gigantic fuckup. I don't want to do it Microsoft's way, I want to do it the right way, the standard way, the correct way.
What will HTML5 fall back to when a non-HTML5 compatible browser is being used?
Fallbacks are an orthogonal and unavoidable issue. Really, the only media format that'll really play from a browser on 99% of PCs today is a http://www.foo.mpg/ link to a MPEG-1 file. It's great to fantasize about "one format to rule them all" but even when that's introduced in a product, it won't become universal until there aren't any systems left that can't play it. And the best way to embed video will be well beyond HTML5 by the time HTML5 may be universal. Heck, if the baseline is just Theora + Vorbis, it's already 5-8 years behind the state of the art. And we'll proabably have H.265 in 4-5 years...
Do what exactly? Linking off to a site that requires Silverlight with no explanation doesn't seem like a very good argument when you're posting to a forum that doesn't want to install your plugin. (Or more to the point, many of them can't install Silverlight 2.0.)
The second link is about how the stuff in the first is authored, and doesn't require Silverlight. Sorry if that wasn't clear.
Your second link talks about multi-bitrate encoding. Which strikes me as (like the entire Silverlight platform) a solution looking for a problem. Despite the fact that Microsoft has had the technology deployed for years as part of WiMP, the market hasn't bought into it. It's just as easy (and probably less confusing) to simply provide different sizes. 95%+ of current streaming videos don't even have to worry about that.
The challenge with offering multiple sizes is that in forces the user to know what their system and connection can play, and it really only works with progressive download models, not real instant-on, easy random access long-form "streaming." That's fine for some audiences, but not for the mass market. Multibitrate done right means nearly instant startup and gapless playback, dynamically adjusted to what the user's machine can play back. It's a very different use model than YouTube.
Perhaps Silverlight will be better positioned when HD streaming becomes the norm. More likely however, is that HD will be the norm when the majority of hardware on the market is both capable of HD streaming and integrated into the standard home in a way that would make HD streaming a superior enough experience for consumers to want to use it.
Ah, that's the point! Smooth Streamings gets us out of having to wait for everyoen to be able to do HD to use it for mass audience content. If only the top 40% of users can get full 720p, the top 40 % of users get full 720p. And users who have less get the best experience their hardware and network is capable off. We don't have to sweat the lowest common denominator.
As an aside, why is it that every Silverlight website stops you cold? There's not even a description of what it is you're missing and/or why you should install the plugin. It's simply "install this or go away". So I go away. No skin off my nose.
There's a lot of flexibility in how a site can present the install option. For example, NBCOlympics.com offered a fallback to an IE embedded WMP ActiveX component. I agree that more sites could do a nicer job of it, and we're talking to them about improving that experience.
And I don't see anything in Silverlight that isn't similarly addressed by HTML5. Ergo, HTML5 is superior for its standardization, true cross-platform support, and competing implementations that can meet the needs of many different ideals.
As I said elsewhere, HTML5 has no way to do anything like this:
http://www.smoothhd.com/
http://on10.net/blogs/benwagg/Expression-Encoder-2-Service-Pack-1-ndash-Intro-and-Multibitrate-Encoding/
As for cross-platform, Moonlight 2.0 should be able to run SmoothHD just fine, and more importantly a whole lot of content published using that platform.
Moonshine runs Moonlight from inside a Firefox plugin to emulate the older Windows Media Player ActiveX embeds. It can also play back local WMV files.
http://abock.org/moonshine/
Which is good, as VLC and thus presumably ffmpeg hasn't been able to play VC-1 files with B-frames for years, which is pretty much the default these days.
I don't see anything in that demo that wasn't in Silverlight 1.0 demos a couple of years ago.
On the media side, check out:
http://www.smoothhd.com/
I encoded the "Big Buck Bunny" clip up there :). It's still in pre-alpha, but you should be able to get the idea
This uses a new API called MediaStreamSource, which enables file parsers and protocols to be built in managed code, and then hand off the video and audio bitstreams to Silverlight's built int decoders.
In the case of Smooth Streaming, every two seconds of the video is a seperate http request, and each of those chunks is available in six different data rates. Managed code heuristics running inside of Silverlight dynamically pick the right bitrate for the next chunk based on available CPU power, network speed, and window size (no reason to download 720p if the brower window is shrunk down in a corner of the stream).
And because this is based around small http requests, chunks get proxy cached, so 100 people watching the same video behind the same firewall would only need to get a single copy, providing much better scalability than traditional unicast streaming.
Anyway, this is something that Flash certainly can't do, and I haven't seen any hint of HTML5 being able to do. Pulling it all together requires some pretty specific characteristics of the video decoder (the ability to switch resolutions with a new sequence header without any pause), an API like MediaStreamSource, and having a performant enough runtime to be able to run all the heuristics and parsing without using much CPU.
I blogged the authoring workflow for this and some other details here:
http://on10.net/blogs/benwagg/Expression-Encoder-2-Service-Pack-1-ndash-Intro-and-Multibitrate-Encoding/
Nope, not yet. Netflix requires Silverlight 2.
But when you look at it any more than superficially, every dollar the government spends is taken at gunpoint (perhaps metaphorically) from someone.
The difference between being threatened with execution and having ones wages garnished is rather more than metaphorical! A just society can and should be financed without shooting anyone's grandmother.
Even incarceration is reserved for those who file fraudulent taxes, not those who fail to pay for one reason or another.
No it doesn't; it increases the profit margin of the people providing the goods. Why on Earth would they lower prices when they're making good money? Demand is not perfectly elastic in the real world (especially for necessities like food, housing, health care, etc.), so lowering the price won't necessarily make a seller more money by increasing the amount demanded. Therefore, it's in their best interests to just pocket the money, and that's exactly what they do.
Interesting theory.
Fortunately, we have rather a lot of empirical evidence on this topic! For your testable hypotheses:
Hypothesis 1: The price of imported goods imported from a low cost country will not go down.
Results: They in fact went down like crazy. Have you seen what a kid's $20 toy can do these days? Shopped for discount shoes? You can buy decent quality stuff for much cheaper in inflation adjusted dollars than you used to be able to.
Hypothesis 2: Earnings of companies who largely import good will go up in proportion to the savings of manufacture in low-wage countries, less the copy of transport.
Results: Nope. US corporations have pretty consistant average profits. This is predicted by economic theory. if a 5% return on capital is considered "good enough" by the market, and a company figures out how to get a 50% return on capital one year, every company in the same industry would try to copy the same technique and would be very happy to get only 20% (4x what they had before) in order to get market share, until return on capital eventually gets back to 5%. The way to be hugely profitable is to do something no one else can do, not to do what everyone else is doing a little cheaper.
Couple that with the enormous capacity increases you're seeing in regular MP3 players, and there's not much reason to go for Ogg anymore. Encode your stuff as 192kbps MP2, and it's future proof, playable on free players, playable on virtually every portable player, and higher quality than MP3 at the same bitrate. Go figure.
Er, NO. While perhaps MP3 @ 384 Kbps can outperform MP3 @ 320 Kbps (never tested that myself), MP3 mainfestly outperforms MP2 at the data rates where you can hear a difference. The old rule of thumb is that MP3 offers about 50% better efficiency than MP2, so that 192 Kbps MP2 will sound roughly like 128 Kbps MP3. Which I don't find high enough for music listening, particularly in CBR.
I don't know if I should laugh or cry. On the one hand, $100,000 is serious money. On the other hand, it barely pays for a good developer for one year.
A good, experienced encoder optimization engineer would cost a lot more than that as a contractor or as an employee ($100K salary, maybe, but a whole lot more with benefits, overhead, and decent quality gear to evaluate video quality with).
The challenge for Theora is to get competitive "enough" ("enough" being specific to use ) versus H.264 and VC-1 not as they are today, but as they'll be in the timeframe of those improvements. There's a whole lot more then $100K being spent by a variety of companies and groups to squeeze improvements out of the commonly used standard codecs as well.
Theora is a weird enough of a design I don't have a clear intuition of how good it could be. Lacking bidirectional prediction (B-frames) is a pretty big limitation out of the gate. But there's no doubt lots of room for it to be improved compared to where it is now.
Yes. Since a game is almost the canonical exclusive mode task, the OS should add as little overhead as possible, with as many cycles as possible going the the CPU and GPU.
If one version of Windows had a lot higher performance than another for gaming, that'd be more suggestive of a bug or driver issues in the slower version.
It's an easy thing to fix - require that H1B visa holders receive the same pay and benefits for their work as the rest of the workforce. If companies really have problems finding citizens to fill jobs, and aren't just trolling for lower paid wage slaves, then it ought not to be a problem, right?
Working at Microsoft, this is how it's always seemed to me. I've worked with some people for serveral years before finding out that their immigration status isn't what I'd assumed. Beyond some people having to go wait in line at the DHS for paperwork ocassionaly, I haven't noticed any practical distinction based on status. And I'd consider it rude and inappropraite if I did.
I certainly agree with your proposal, but that's essentialy what's happening already from my experience at Microsoft from what I've seen.
And there's absolutely technical sub-disciplines where there simply aren't enough US graduates to fill open positions. From my own world, signal processing is a big one. We see a roughly even mix between North America, greater China, ex-USSR, and ex-Raj origins for people who work on encoder and decoder stuff. Lots (most?) of the people who have been around for long enough have become citizens.
If anything the right solution would be to just not have a cap for skilled workers so that a seperate H1B program isn't needed. This is how Canada does it (you've got the right skills, etcetera, and you can come), and they're a better country for it.
Read Peter Gutmann's excellent article here:
http://www.cs.auckland.ac.nz/~pgut001/pubs/vista_cost.html
Why are we still talking about this? Gutmann had never run Vista when he wrote the article, and stopped updating it once Vista was in the market and actually testing it was possible.
He makes a ton of testable hypothoses in the article. And I don't see a one that hasn't been disproven.
It's simply not an article about Vista. It's an article describing his fantasy of how bad Vista could have been.