Fast Wi-Fi's Slow Road To Standardization 140
Posted
by
timothy
from the vested-interests-took-off-gloves dept.
from the vested-interests-took-off-gloves dept.
CWmike contributes this excerpt from Computerworld: "For a technology that's all about being fast, 802.11n Wi-Fi sure took its sweet time to become a standard, writes Steven J. Vaughan Nichols. In fact, until September 2009, it wasn't, officially, even a standard. But that didn't stop vendors from implementing it for several years beforehand, causing confusion and upset when networking gear that used draft standards from different suppliers wouldn't always work at the fastest possible speed when connected. It wasn't supposed to be that way. But, for years, the Wi-Fi hardware big dogs fought over the 802.11n protocol like it was a chew toy. The result: it took five drama-packed years for the standard to come to fruition. The delay was never over the technology. In fact, the technical tricks that give 802.11n its steady connection speeds of 100Mbps to 140Mbps have been well-known for years."
Re:Drama...? (Score:5, Interesting)
I went through two small companies working on pre-standard N devices. Both went under as a little company you can't pre-run a standard to market. We were ready for production 7 years ago.
So yes, drama that personally affected me as I went through two collapsing companies.
History Repeats Itself (Score:5, Interesting)
Re:Unimpressed with 802.11n (Score:4, Interesting)
Basically any 802.11n system is going to be shit with legacy devices nearby. I believe this was one of the final barriers to standardization between Draft-N and final.
So to get decent performance you MUST be in the 5 GHz range. However, there are almost no 5 GHz 802.11n routers out there with external antennas! (Curse Linksys for their move to "saucer" form factors across their entire 802.11 product line... 100% driven by form and aesthetics/marketing, not technical function.) End result is that in most situations, a G router with an external antenna (especially an upgraded one) will blow nearly any of the 11n devices on the market out there.
Re:Not the first time (Score:3, Interesting)
I've been running wired GigE for 4 years.
I also have fiber in the same cable strand, so I could go A WHOLE LOT FASTER.
So your lame bragging doesn't really sound so hot.
Wireless has always been a train wreck compromise for people too cheap to set up a proper network.
Blame the Manufacturers (Score:2, Interesting)
Despite the moniker of "open standard" every vendor who contributes to these standards and who has "voting" authority on them have to maintain their business interests. 802.11n was held up more for business reasons, members are competitors remember, where some didn't have product available. They obviously want to make sure that their engineering and pre-manufacturing ramp ups are in line before the standard is released. Like 802.11n, this didn't stop many vendors from releasing "pre standard" products as soon as the RF standards were put into place. In reality it then becomes a firmware or driver issue to become compliant once the status is released.
If the standards boards were truly "open" then they'd get the standards drafted, agreed to and voted on in short order. The reality is that they need the industry experts and those experts also have to maintain their company's interests. It won't change, just learn to live with it.
The tricks: LDPC codes. (Score:4, Interesting)
One of the tricks is low density parity check codes (LDPCC) which are the best currently known error correcting codes. They're decoded with a wonderfully elegant decoding algorithm which is embarresingly parallel so it works very well in hardware.
In fact, you can pretty much implement the belief network in hardware directly.
The codes are also used in 10G Ethernet, too.
Funny thing is that they date from the 60's, but were impractical because of the amount of computation required to decode them. The decoding algorithm was then rediscovered for inference on Bayes nets.
If you lick this sort of thing, it is worth reading Mackay's book on inference which is free online. I have no affiliation to Mackay, btw.
Re:Blueray of Wifi (Score:4, Interesting)
Re:Lack of Demand (Score:4, Interesting)
1 GB file
802.11b- 27 Minutes at 5 Mbit/sec (625 KB/sec)
802.11g- 8.5 Minutes - 7 Minutes at 20 Mbit/sec (2.5 MB/sec)
802.11n- 1 Minute, 15 seconds at 110 Mbit/sec (13 MB/sec)
100-BaseTX- 1 Minute, 30 seconds at 92 Mbit/sec (11.5 MB/sec)
1000-BaseT from/to Laptop drive- 17 seconds at 480 Mbit/sec (60 MB/sec)
1000-BaseT from+to high-performance desktop drive- 11 seconds at 800 Mbit/sec (95 MB/sec)
1000-BaseT RAM --> RAM - 9 seconds at 945 Mbit/sec (118 MB/sec)
60 MB/sec is realistic to expect when transferring to or from a laptop with a 5400 RPM drive. 85-95 MB/sec or even 100 MB/sec+ is achievable when transferring between high-performance 7200 RPM desktop drives, at the beginning of the drive.
However, 1 GB is small. A typical HD tv show is 1.1-1.4 GB. A typical 720p x264 encode is 5-8 GB. A typical 1080p x264 encode is 8-15 GB. A system backup can be anything from > 10 GB for incremental backups to 60-100 GB for full backups of system partitions.
At 85 MB/sec you can transfer a DVD (4.37 GB) in 60 seconds.
Re:History Repeats Itself (Score:3, Interesting)
Indeed it does.... and you don't know the half of it.
IEEE operates with a completely different dynamic from what most internet folks are used to.
One of the big motivations for a company to sponsor a participant (an engineer, by paying him to prepare and to attend) is to get the company's intellectual property incorporated into the standard under development as a MUST. This is all above board, and the companies must declare up front if they believe they have IP in a proposal and to agree that if adopted they will license the IP to implementers at "fair and reasonable" rates.
When there is only one proposal, or when one is clearly superior, live is good and things typically move along smoothly.
On the other hand, when there are multiple proposals with relatively equal technical merit, it can drag out.
This happened in g, and in again in n.
We narrowly headed it off in i.... At one point the, the AES cypher in the draft as a MUST was OCB (Offset Code Book), and incumbered by no less than three independent patents. This had two implications that several of us strenuously objected to:
- First, implementers would have had to license three times to be certain they were in the clear, thus increasing the cost of the chips and the end product.
- Second, since the open source community has no real way to execute said license agreements or to pay royalties, this would have guaranteed that it would have been impossible for there to be a "legal" open source implementation.
In the end, we prevailed and the AES cypher in the spec is CCMP, which is not encumbered.
Sadly, this is just the way it is..... at least in -some- standards organizations.
Red
Re:Not the first time (Score:3, Interesting)
It's funny you mention Flash.
It's one of the few things that really "just works" across all browsers, regardless of the underlying O/S. Perhaps that is why the web as a whole has adopted Flash so readily, (for better or worse), and why I feel we'll never be rid of it even when html5 is "live".
Now it's proprietary, which means the owner gets the control over what the user can see or not see (source), but let's face it, it doesn't take Adobe 5 years to bang out an even better version, what are we on now, version 11 or 12 ?
If, and I know it will never happen (just hypothetical ponderings), Adobe open-sourced it tomorrow and released the complete data format specs, so that the underlying instructions and objects could be expressed in say XML, do you think we'd need html5, or indeed html markup at all, ever again ?
Just trying to think what you could do with html5 that you couldn't already do with flash where the applet / player read from an XML "data" file that anyone could edit without the need for proprietary IDE's and design software etc.
The more I see of html5, the video, canvas and drawing support we've been missing for so long, the more it looks like a SWF object animation format.
Re:Drama...? (Score:2, Interesting)