First, WinModems. Now, WinWiFi.

zulux writes "Microsoft is actively encouraging WiFi (802.11b) hardware manufacturers to strip their devices of costly electronics, and use Microsoft software/drivers to make up the slack. And you thought WinModems were bad!"

Comments

Re:Bias, bias, bias

[Toraz Chryx]

I've got a Diamond Supra 56k winmodem laying around here..

it's performance is... well, crappy.
300ms pings and 3.5KB/s transfer rates != good

and before you say 'well, that's your phonelines'
a USR sportster external manages 150Ms pings and 5KB/s+ transfer rates plugged into the same socket.

both are fairly moot now, thanks to my 'wonderful' software based Alcatel SpeedtouchUSB (didn't have much choice there, they wanted a higher monthy fee for an ethernet connected DSL package.. and they weren't doing 'we provide the line, you provide the hardware' at the time)

802.11a already has this problem

[billstewart]

I've been told by somebody working at an 802.11a manufacturer that the specs for that are designed to have significant parts of the system run on the host computer's CPU rather than the card, and that therefore there'd be issues with getting Linux drivers unless the manufacturers funded them.

Some of the concerns are the amount of processing horsepower required for security and maybe also for some of the communications functions, since it's easier to add computational horsepower when you're not crammed into a small card competing for space and heat load with the radio circuitry, and also convenience in upgrading the system, especially if upgrades may require even more substantial increases in CPU crunching, such as bigger RSA modular multiply/exponentiations.

The importance of convenient upgrades has been amply demonstrated by the repeated failures of WEP :-)

== Apple's "Software Base Station"

[rheiser]

This sounds suspiciously like the "Software Base Station" available on Macintoshes for a number of years now (surprise, surprise!) It allows you to use a computer with an AirPort card to act as a Base Station for other computers with AirPort cards, instead of spending the $250 to buy a dedicated one.

rheiser

Wrong Interpretation?

[__aadidx2690]

Judging from what I've read elsewhere, the submitter may have interpreted the article a bit wrong. It's not so much that MS and Intel (also mentioned in the article) want to have the WinModem equivilent of 802.11, but that they want to make the access points cheaper by providing a software solution.
Apple has had a similar product, the "Software Base Station," available for Mac OS 9 for quite some time!
See this (much better) article for details.

Future SoC Design

[LoRdTAW]

My first reaction to this topic was "Oh god here we go again" but then I sat back and thought a little.

First think of the popularity of CNR and AMR sound, ethernet, dsl and modem cards you have seen for sale. Answer none. It just never took off because noone wants to sell cheap hardware when they can make more money with a real hardware solution. At one point they litterd every intel bord I ever came across and then only appeared on cheap bargin pc motherboards. I have been hearing about soft dsl for a long time and I have never seen a soft dsl card, also all soft dsl cards support the G.light standard which no provider cares about.

Second face it, There is alot of money to be made with hardware. The 3D graphics market took the exact opposite approach to this problem. the first popular true 3d polygon game was quake. now the old dos quake ran in total software, everything from the AI to 3d graphics was done on your cpu (back then I ran it on a 486 with a p83 overdrive) and now look Nvidia took the whole graphics pipe and threw it on a chip which is totally opposite the software approach. Some people would sell there own mother for a geforce 4 if they could. Shure today the cost of CPU's have come down enough to justify the $400 tag on a GF4 Ti but take away that pipe on a chip and do it in soft. People would scream bloody murder. And sound cards are going this way too. Pretty soon we might be shelling out 200+ for a sound card with a APU (audio processing unit) that will imerse you in a whole new world of sound

And last relax people with SoC (System on Chip) design coming along nice these days I wouldent worry. There are already all in one SoC's for DSL modems and cable modems. I imagine a cheap WiFi solution is in the works as we bitch.

Re:Using CPU cycles can only go so far

[MtViewGuy]

Hmmm... with all that signal processing oriented compute power on the graphics card, why not make that available to do the wifi demodulation or software radio in general?

One good reason: simplicity of computer architectural design. Putting the decoding process into the CPU keeps the hardware count down, for starters. That is the reason why on the x86 compatible side we've seen the addition of CPU registers oriented towards multimedia processing: Intel's MMX, SSE and SSE2 and AMD's 3DNow! and 3DNow! Professional.

In the case of decoding MPEG-2 video streams from a DVD movie disc, you can decode them in software pretty reasonably well, especially with today's 1,000 MHz and faster CPU's. However, that still means using a lot of CPU cycles doing it, and that means other programs may start to drag because the CPU isn't so available.

Due to the fairly computationally-intensive process involved in decoding MPEG-2 video, that's why there has always been interest in off-loading the decoding process somewhere else. That's why when DVD-ROM's first started showing up on PC's we saw separate decoding adapter cards from Creative and Sigma Designs so the decoding is completely done by these specialized cards. When CPU speeds got fast enough and Intel introduced the wider bandwidth AGP connector, ATI implemented HWMC and IDCT assistance for MPEG-2 decoding on the graphics card itself starting with the Rage 128 chipset, which off-loaded most of the MPEG-2 decoding process from the CPU; most other chipset manufacturers (S3, SiS and nVidia) soon had at least HWMC assistance. The success of ATI with this way of MPEG-2 decoding is the reason why the nVidia GeForce4 MX and GeForce4 Ti series of chipsets now have multiple levels of hardware assistance for MPEG-2 decoding, not only for playing back DVD movies for also eventually for playing back 1080i 16:9 HDTV video.

some of the brave of us attempt to reverse- Ah ah.

[crovira]

Naughty naughty...

There's the DMCA (or whatever floated to the top of the "Alphabits"TM bowl of some congressman's breakast that morning,) to slap you down with if you even try that.

Don't you know yet that YOU have no rights?

WinWiFi != WinModem

[Pizza]

As someone who spends a considerable amount of time these days hacking on WiFi card drivers, Host-based MAC is actualla a VERY good thing.

A good analogy of this is PPP. The current situation is similar to a modem manufacturer embedding PPP in the hardware, which is horribly complex and expensive to implement. It is much simpler and cheaper to let the OS provide the PPP services.

WinModems come in two flavors; host-based controller and host-based signal processing. The latter is pure evil; the hardware is nothing more than a A/D/A converter, and the host CPU has to perform all DSP functions to make it into a modem. The host-based-controllers have real hardware DSPs and whatnot, but they just tell the DSP what to do, essentially replacing an on-board processor+firmware with the driver on the host machine.

WinWiFi (which is really host-based-MAC) is neither. The WinWiFi card would become about as smart as the average ethernet card; ie it would be able to transmit and receive raw 802.11 frames, and then pass them off to the driver which then figures out what to do with them.

A good portion of the wireless cards out there already do this, and nearly all of the new ones will do this. Why? complexity and cost.

802.11 is rather complicated. The MAC must handle a complex state machine; with all sorts of little nuances. Handling transmits/receives, and their acks, association, channel hopping, and then the real doozey: encryption.

WEP sucks. Not just because it's fundamentally broken, but because it takes a bit of oomph to work with, and it's a little complex. And if this is done in hardware, you can't update it to handle newer standards.

Every single one of the 802.11 extensions to replace/augment WEP will require considerably more computation power in hardware; but in fact, most 802.11 (windows) drivers now do WEP on the host, because it has far more computational power to spare with zero additional hardware cost.

This WinWiFi initiative is nothing more than "hey, all of you guys have already written this host-based-MAC stuff (or are going to have to write it anyway) so why not just use the stuff already part of the OS? It's already been extensively tested and that way, you don't need to reinvent the wheel."

It's called shared code, and makes a lot of sense.

I've been banging my head against the wall a lot lately because of buggy firmware in WiFi cards; If they let the host OS do the work instead, these bugs wouldn't exist, because the 802.11 spec is well-documented.

And again, it's not WinWiFi, it's Host-based-MAC. It's a work-in-progress for Linux too. And it is a GoodThing(tm).

- Pizza

What are you afraid of?

[Gumber]

What is the chipset used in most WiFi cards? The Prism 2/2.5.

Who makes it?
Intersil

Are there linux drivers?
Yes! With full source!

And guess what, Intersil comissioned the drivers!

Not only that, but the drivers offer support for advanced functions typically not offered on Windows based PCs (host based access point support).

So, based on past history, there seems a good chance that there will be a path to Linux support for WinWi-Fi cards.