Slashdot Mirror


Intel Putting Wi-Fi into Future Chipsets

Ridgelift writes "Wired's got the story on Intel's plan to incorporate Wi-Fi into the motherboard chipset. "The chipset, however, will not include an actual Wi-Fi radio, so users will still need a wireless add-on card. Intel has said it eventually intends build a Wi-Fi radio into its microprocessors." This would make setting up a wireless network a lot simpler."

51 of 216 comments (clear)

  1. Is wifi on CPU a good idea? by brejc8 · · Score: 5, Insightful

    It's all very well putting more peripherals onto processors but with the shrinking feature sizes and an increase in cross talk is placing a powerful 2.4GHz source on die really a good idea? You would have to shorten long lines and slow down the processor or suffer random errors. Doesn't inspire me with confidence.

    1. Re:Is wifi on CPU a good idea? by IAR80 · · Score: 3, Interesting

      It is not on the cpu, but the chipset and the RF part is not there. It will just be something like the 802.11 asic dealing with the the protocol not with power up/down conventers or anything like that.

      --
      http://ebgp.net/ccc/
    2. Re:Is wifi on CPU a good idea? by John+Hurliman · · Score: 2, Insightful

      Putting WiFi technology on the motherboard seems like the next logical leap, but hopefully they keep the RF transmitter hardware outside the box. Otherwise say goodbye to HiFi soundcard quality.

  2. Like Microsoft... by Jake73 · · Score: 4, Insightful

    This is a lot like Microsoft's business plan, but at the hardware level. It leaves ISV's out of the loop. "Centrino" was perhaps the most serious example of this that I've seen.

  3. Intel linux support sucks by pe1rxq · · Score: 5, Interesting

    They will probably promise to provide Linux driver like with the centrino chipset and then not even make specs available.
    You will get all kind of lame excuses:
    - We are working on a driver.... (For half a year already)
    - We can't tell you how to operate it because the FCC won't let us (Complete bullshit but sounds nice: 'linux hackers want to interfere with police radio')
    - They might release some binary only modules... (Redhat version bla.bla, kernel version bla.bla and nothing else)

    Jeroen

    --
    Secure messaging: http://quickmsg.vreeken.net/
    1. Re:Intel linux support sucks by Homology · · Score: 5, Informative

      Intel are not forthcoming with documentation about NIC and crypto. Here is what OpenBSD has to say about this
      OpenBSD experience with Intel :


      Much like Intel does for all their networking division components, and completely unlike most other vendors, Intel steadfastly refuses to provide us with documentation. We have talked to about five technical people who are involved in the development of those products. They all want us to have documentation. They commend us on what we have done. But their hands are tied by management who does not perceive a benefit to themselves for providing documentation. Forget about Intel. (If you want to buy gigabit ethernet hardware, we recommend anything else... for the same reason: most drivers we have for Intel networking hardware were written without documentation).

    2. Re:Intel linux support sucks by gmhowell · · Score: 2, Informative

      Not sure if asx100 and acx100 are the same thing, but if so, have you tried this?

      --
      Jesus was all right but his disciples were thick and ordinary. -John Lennon
    3. Re:Intel linux support sucks by O · · Score: 2, Informative

      Um, I know Dell sells them. There was a 802.11b MiniPCI card from Dell on sale for $29 the other week. Check Techbargains for it. Also, if you're desperate, you can generally rip apart a consumer AP or "router" and yank out it's MiniPCI card, too.

      --

      1, 1, 2, 3, 5, 8, 13, 21 -- Mathematics is the Language of Nature.
    4. Re:Intel linux support sucks by gordyf · · Score: 2, Informative

      I know you said retail, but I bought a lucent orinoco minipci card off Ebay a little while ago. $35 for the minipci card attached to a real pci card - removed two screws, popped it out of the socket and replaced my Intel pro/2100 (piece of crap), and have been happy ever since!

      It uses the Orinoco cardbus drivers, since the orinoco minipci card is essentially a PCI card with a TI 1410 PCI/CardBus bridge and a CardBus orinoco card all in a minipci form factor. Cool stuff.

    5. Re:Intel linux support sucks by pe1rxq · · Score: 2, Insightful

      However intel have publicly stated that they would have linux support for their centrino chipset...
      (which the pro wireless 2100 is part off)
      That turned out to be one big lie.

      Jeroen

      --
      Secure messaging: http://quickmsg.vreeken.net/
    6. Re:Intel linux support sucks by megabeck42 · · Score: 2, Informative
      Frankly, this is taken out of context and is rather misleading. Sometimes I wish for a moderation option entitled "Wrong."

      First of all, the section above is listed under the header called "Intel Ipsec Cards" and more specifically refers to the Intel Encryption Coprocessor on the card.

      Further, Intel has written and released a free, GPL ethernet driver for their EEPro 100, 1000, and 10000 ethernet cards. I shall transcribe for your benefit the top few lines from linux/drivers/net/e100/e100_main.c:

      Copyright(c) 1999 - 2003 Intel Corporation. All rights reserved.

      This program is free software; you can redistribute it and/or modify it
      under the terms of the GNU General Public License as published by the Free
      Software Foundation; either version 2 of the License, or (at your option)
      any later version.

      This program is distributed in the hope that it will be useful, but WITHOUT
      ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
      FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
      more details.

      You should have received a copy of the GNU General Public License along with
      this program; if not, write to the Free Software Foundation, Inc., 59
      Temple Place - Suite 330, Boston, MA 02111-1307, USA.

      The full GNU General Public License is included in this distribution in the
      file called LICENSE.

      Contact Information:
      Linux NICS
      Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497

      That driver is a GPL implementation meaning that the OpenBSD developers are more than welcome to port it at their leisure.

      Oh, you want real documentation too. Take a look at developer.intel.com:
      Intel 82551ER Fast Ethernet Controller Networking Silicon Datasheet
      Intel 82559ER Fast Ethernet Controller Networking Silicon Datasheet
      Intel 82559ER EEPROM Map and Programming information

      True, Those are for their FastEthernet chipsets, not the Gigabit chipsets that, "Intel steadfastly refuses to provide us with documentation"

      Well, what about these?
      Intel(R) 82541ER Gigabit Ethernet Controller Networking Silicon Datasheet
      Intel(R) 82547GI(EI)/82541ER EEPROM Map and Programming Information Guide

      Ok, so the IPSEC chip on the NIC isn't supported nor is there any data on that chip forthcoming. However, there are a number of papers that show that IPSEC and TCP offloading (not to be confused with TCP fragmentation/checksumming) are not efficient. Specifically, the "hardware" IPSEC is done by firmware downloaded to a small embedded processor on the NIC. This small, embedded NIC is not very fast, in fact, its rather slow.

      Result:
      Processor utilization drops marginally (modern processors can encrypt 10 megabytes/s trivially)
      Latency shoots up (It takes the embedded processor longer to encrypt a packet than the host processor would.)

      There are a number of papers corroborating that latency has a huge effect on maximum bandwidth.

      [I think the paper regarding TCP offloading not being worthwhile is by Mudge. The IPSEC offloading not being worthwhile is my hypothesis, untested, but I feel logically founded.]

      My point is that IPSEC offloading is not an advantage - it probably was in 200MHz K6 days, but it certainly is not in 2.0 ghz K7 days.

      Other notables, for example, the 3Com 3CR990 still doesn't have IPSEC offloading, despite promises from either the openbsd txp driver or the linux typhoon driver.

      Frankly, as far as Gigabit

      --
      fnord.
  4. Really? by odyrithm · · Score: 5, Insightful

    This would make setting up a wireless network a lot simpler.

    Woah there, whats so hard with the way its setup now? pcmcia is a matter of plug and go, pci is a matter of modprobe if that.. theres nothing hard about wifi... its a nic with a wireless medium.. thats all.

    Now if intel had some new fangled wep replacement then that would make things simple, no more mac rules on my fw would be nice.. which is unlikely.

    --
    moo
    1. Re:Really? by Artifex · · Score: 2, Interesting
      Woah there, whats so hard with the way its setup now? pcmcia is a matter of plug and go, pci is a matter of modprobe if that.. theres nothing hard about wifi... its a nic with a wireless medium.. thats all.


      You guys are not looking at the big picture, here. Adding this on board allows OEMs to specify a motherboard platform with this, and then buy the cards at cheaper prices than full implementations on cards would cost. It's just following the trend of AMR slots and onboard video, and, in the last couple of years, wired NICs.

      Early hard drive and floppy controllers used to only be on cards, too.

      --
      Get off my launchpad!
    2. Re:Really? by be-fan · · Score: 2, Interesting

      I'd say the number is higher than 99%, but its exactly that few percent that are the kicker. With modprobe, I can type "modprobe i810_audio" and the thing just starts working. If something goes wrong, I can check the dmesg output, and ahve a range of debugging options. With Windows, if something goes wrong, you're screwed. I've been trying to debug persistent wonkiness in the wireless card of one of my Windows machines. After months, I've just given up and take to reinstalling the drivers every month. It doesn't help that there are two competing wireless managers (the one that came with the card, and the one that comes with Windows) and I really don't think I can reproduce the exact steps I took to disable one and keep the two from interfering with each other.

      Of course, that's largely a moot point --- current distros use something like hotplug or kudzu, so you don't have to use modprobe anyway. But the advanced troubleshooting options are still there if you need them.

      --
      A deep unwavering belief is a sure sign you're missing something...
  5. What's the point? by bobthemuse · · Score: 5, Insightful

    The chipset, however, will not include an actual Wi-Fi radio, so users will still need a wireless add-on card. Intel has said it eventually intends build a Wi-Fi radio into its microprocessors.

    Why put in a chipset without a radio? Maybe one could argue an attempt to get market share by making their add-on card cheaper than the others (just radio, no chipset), but this card will have a more limited market, since it wouldn't be compatible with older or non-intel mobos.

    Now if the were to put a software radio on board, *that* would be cool! Think of upgrading to future standards with just a flash rom upgrade...

  6. users will still need a wireless add-on card by rokzy · · Score: 4, Insightful

    I don't follow tech too closely but this sounds like those worthless AMR slots or whatever they were called - the ones that were like having a built-in sound card or modem (in the sense that the board cost more and had less space available) except didn't actually do anything...?

  7. Dumb Question by Anonymous Coward · · Score: 2, Funny

    Customer:If I can have a 3 Ghz processor why cant I have WiFi at 3 Ghz
    Radio Shack Assistant: Mam the 3Ghz speed refers to CPU speed while Wi Fi works at 2.4 Ghz
    Customer:Dont get technical on me . I know this Ghz speed keeps increasing all the time . It cant be fixed at 2.4 Ghz

  8. Hospitals and Airlines beware by ericspinder · · Score: 2, Interesting

    Building more functionality into the motherboard is an ongoing trend, but adding a radio cannot be a good thing. Due to potential interferance, you cannot go into a hospital or airplane without being told to turn off your cellphone.

    --
    The grass is only greener, if you don't take care of your own lawn.
  9. Wouldn't this compromise security? by twiztidlojik · · Score: 4, Insightful

    I think that if an admin of a corporate network had several thousand wireless clients, a simple airsnort would compromise thousands of computers, without requiring the packet sniffer actually access any sort of ethernet cabling.

    This could have a grave impact on the sales of wireless-based chipsets in the corporate market.

    --
    I will now redundantly add my name to the end of my post. You know, in case you forgot me or something.
  10. how convenient by vinsci · · Score: 4, Insightful

    So when Big Brother wants to know what you're up to, they don't need you to be online to the public internet. How convenient. The "trusted" BIOS can always let them bypass your firewall, as the BIOS is going to handle the net connection too.

    --

    Trusted Computing FAQ | Free Dawit Isaak!
  11. Unintentional network compromises through bridging by AgTiger · · Score: 5, Insightful

    I sincerely hope they're going to have these things configured in one of the following manners:

    1. Disable the Wi-Fi by default
    2. If not disabled, seed the encryption key with a pseudorandom number before the user specifically configures it.

    You don't want new computers forming unintended bridges or access points between the untrusted network/airspace and your trusted internal network between when they're first powered up and when the overworked sysadmin has a chance to configure them properly. So much for your company's firewalls having a chance to do their job.

  12. Huh? by Aneurysm9 · · Score: 2, Insightful

    Wi-Fi without a radio? Sounds kinda like a soundcard without a DAC or a videocard without a RAMDAC. What's the point of including functionality while not including necessary pieces except, perhaps, to rachet up the marketingspeek and pressure out other manufacturers?

    --
    There was Cowboy Neal at the wheel of a bus to never-ever land.
  13. Should wireless really be in the chipset?? by Anonymous Coward · · Score: 3, Interesting

    I think it should be integrated into the motherboard (like regular ethernet) but not the chipset.

    Or is intel upset that sales of centrino are so poor? To be an "official" centrino laptop, you have to offer intel's 802.11b wireless. Not surprisingly, many people want faster (802.11b/g or 802.11a/b/g) wireless cards.

    Broadcom has been eating intel's lunch in the oem ethernet (wireless & wired) card market. Sounds like anticompetitive monopolist activity to me.

  14. WiFi Security by Chalybeous · · Score: 4, Interesting

    What about security issues?
    I like my router (at home, I share a cable internet connection between two desktop PCs and, occasionally, my laptop). It has an inbuilt firewall according to the manufacturer, and I know if I ever do have a serious problem which I suspect originates from the internet, I can physically disconnect it. Sure, cables are archaic, but they're cheaper and more secure than wireless networks - especially for the novice (like myself).
    But if you enable a CPU to act as a wireless hub - or, eventually, if WiFi comes as a full onboard feature (rather like many motherboards now have onboard sound and graphics) - would that not open up your PC and network to security issues? My parents would not be best pleased if someone warchalked on the fence, but since they have little idea of technology or computer security, I think if they bought a new machine enabled with this kind of tech, every l33t hax0r in a two mile radius would be camping out to leech their access.

    Any other thoughts, opinions or predictions?

    --

    "It is dark. You are likely to be eaten by a grue." -- Zork

  15. Not complete bullshit by mindstrm · · Score: 3, Insightful

    The FCC DOES require that ISM band consumer devices opearate within specifications. They are certified for a given antenna configuration, among other things. Even the anntenna connectors are non-standard, for this very reason (so the consumer doesn't think you are supposed to hook it up to an amp).

    It's nto that you aren't allowed to modify it legally, of course you are, as long as you operate within spec... but that the company has to make it so.

    It's not a big stretch for them to feel releasing driver code is a risk, as anyone who gets it from them could easily use it to operate outside of spec.

    1. Re:Not complete bullshit by timeOday · · Score: 4, Interesting
      It's not a big stretch for them to feel releasing driver code is a risk, as anyone who gets it from them could easily use it to operate outside of spec.
      It's no different than selling cars that can exceed the speed limit. My compliance with FCC regulations is not Intel's concern.
    2. Re:Not complete bullshit by shepd · · Score: 2, Insightful

      They might be able to exceed the speed limit, but how many can go over 250 km/h without being modified? :-)

      --
      If you could be told what you can see or read, then it follows that you could be told what to say or think - BoC
  16. There are other problems by plusser · · Score: 5, Informative

    The reason why Intel probably don't want to integrate the Micoprocessor with the actual WiFi transmitter and receiver is quite simple. If they add it inside the IC, they will have to go through radio use approval for every different potential market in the world, before they sell a single component. By letting the motherboard/add on card manufacturer's do this instead, they can concentrate on developing better microprocessors.

    As somebody in the know, I do worry that these new WiFi enabled equipment could be the next mobile phone when it come to interference of avionic systems; especially as many modern microprocessors are prone to soft faults at altitude due to the effects of the upper atmosphere.

  17. To clarify: by HoldmyCauls · · Score: 2, Informative

    This is to make it so that an average desktop computer can function as a router for WiFi traffic in the home or office. The card is needed NMW, in order to grab that traffic. A poster above mentioned using a software radio, but it seems that that would only be useful if things were reversed: the software radio *interprets* the signal, and *generates* one to return to the WiFi device in question, but ultimately it is a radio device which transmits that signal into the air. The problem Intel will face is explaining this in terms that a PHB who *signs* the check to buy this stuff.

    --
    Emacs: for people who just never know when to :q!
  18. Bluetooth? by ybmug · · Score: 5, Insightful

    802.11a/b/g on board would be nice.. but, I would really like to see more motherboards coming with bluetooth onboard. This would seem to make sense with things like keyboard, mice, headsets, and cellphones that are bluetooth enabled coming to market.

  19. DRM implementation by 4volt · · Score: 3, Informative

    I'm all for WiFi everywhere, but it sounds like a pretty big backdoor to me, I don't think I'd want to have a WiFi connection built onto my board that I couldn't disable with anything mroe then software. Next thing you would know Microsooft is using it to send DRM related information or usage stastics without you knowing.

    I realize that it would probably able to be disabled in BIOS, but it wouldn't take much that if M$ wanted to take control they could do it with a few sentences in the EULA.

    Improbable, but possible.

  20. Re:Why do companies expand into other est business by Artifex · · Score: 5, Insightful
    I think it's silly how Intel is now embracing a trend to include features already provided by another market. Intel will probalby do this witout adding more than 20 employees and in turn drive about 10,000 people out of jobs due to their companies going out of business.


    Oh, yes, it's Intel's fault if 20 of their employees can make a product better suited to the market than 10,000 other peoples' efforts.

    By the same token, buses and trains and taxis have all taken jobs away from the good hard-working people with horse-drawn taxis. And we really should go back to rooms full of seamstresses making clothing by hand, like before those evil industrial looms were created. Oh, and the cotton for the clothes (synthetics put farmers out of work) should be picked by immigrant laborers.

    It's not society's obligation to prop up inefficient methods of production; quite the contrary. There's a word us old-timers use sometimes. It's called "progress." Might want to study it.

    --
    Get off my launchpad!
  21. Easier? by toupsie · · Score: 4, Insightful
    This would make setting up a wireless network a lot simpler."

    How freaking simple can it be now?
    1) Insert Airport Extreme Card into PowerBook 12"
    2) Turn on PowerBook 12" 3) Select Network from Airport menu & Enter WEP if needed
    4) Wirelessly communicat

    --
    Strange women lying in ponds distributing swords is no basis for a system of government.
  22. call me paranoid by ciaran_o_riordan · · Score: 2, Insightful

    I'm no Wi-Fi expert, but couldn't a wi-fi-enabled cpu transmit data without your permission?

    Unique cpu ids? Treacherous Computing Group data?

  23. real transcript by rootofevil · · Score: 2, Funny

    Customer:If I can have a 3 Ghz processor why cant I have WiFi at 3 Ghz
    Radio Shack Assistant: Oh, i dont know. I bet it does.
    Customer: Ok. good.

    --
    turn up the jukebox and tell me a lie
  24. Monopolistic BS by drix · · Score: 2, Informative

    This just is another step in Intel's ploy to rule the wireless market through cheap and underhanded business practices. Not many people know this, or at least I didn't till I started shopping for a laptop 2 days ago, but all new laptops carrying the Centrino designation have to come with an Intel miniPCI WLAN card preinstalled or they cannot be called Centrino. Which is great except that Intel refuses to support Linux on their stinkin' card. (Yes I could go elsewhere, but for the price, speed, and power consumption, Centrino is far and away the best on the market right now.) If you want to monopolize an entire hardware sector, fine--good luck trying. But don't chain me to a stupid Wintel platform because of it. If Intel had their way they'd be the only supplier of WiFi cards within a few short years--then WTF do we do if we're not on Windows?

    --

    I think there is a world market for maybe five personal web logs.
  25. Ack by FrostedWheat · · Score: 4, Interesting

    But why!

    Onboard soundcards (chips?) are rubbish, onboard NICs are quite often crap (not always), onboard modems are a joke and onboard video is nasty. Apart from some specific cases (VIA's mini-itx stuff) I think manufacturers should be moving away from this onboard-everything obsession.

    PCI was invented for a reason! Customisability is what set the PC apart from the Amiga or similar machines!

    1. Re:Ack by antiMStroll · · Score: 2, Informative
      We use on-board ACL 650 sound for commercial use and it's fine, much better than being forced to load Creative's nasty software (circa Audigy 1, haven't tried the 2.) The motherboards we select use Intel integrated NICs, vastly superior to the typical Realtek-level junk found in so many machines. Modems, you have a point, integrated video however has its uses. The desktop demands in a corporate environment are typically meagre. Integrated handles the task fine and has the added benefits of reduced noise, heat and cost, plus increased reliability (no GPU fan) and generally better stability since the GPU and MB were designed as one.

      If AutoCAD or UT2003 aren't part of the design spec, motherboard integration makes a great deal of sense.

    2. Re:Ack by megabeck42 · · Score: 2, Interesting

      Frankly, that's a product of the driver not a result of chip design or flaws. Further, all Windows (NDIS?) drivers I've had to work with have had the option called "Locally Administered Address" where you can force a custom MAC address.

      If you take a look at most ethernet drivers, you'll notice that most drivers need to read a MAC address out of a serial eeprom. This address is then configured into the ethernet chip so it can receive appropriate unicast frames and is used by the driver when constructing ethernet frame headers for transmission.

      Further, regarding just how "junky" the Realtek stuff is, I refer you to this page: http://www.fefe.de/linuxeth/realtek.txt which states quite clearly and specifically why the realtek card is "junk."

      For the record, I can spoof the MAC address on my 3Com 3C905's, Intel EEPro100 and Broadcom Tigon 3.

      --
      fnord.
  26. Re: integration by King_TJ · · Score: 4, Interesting

    Personally, I'm not really convinced all these integrated parts do us any favors.

    Just last week, for example, I installed a new Pentium 4 motherboard and CPU in a standard ATX case that was formerly running a PII system. This was done for a law firm, and was upgraded on-site, because they couldn't afford to have much downtime.

    Well, as luck would have it, the integrated EIDE controller was faulty. I kept getting "data corrupt" type messages when it tried to boot Win2K on the drive that just worked in the other system. I tried a different hard drive with a fresh format, and had the same issue. Even the secondary channel had problems.

    If it hadn't been intergated, I could have simply swapped a $15 or $20 controller card and gotten everything back up and running for them.

    The more devices Intel can integrate into motherboards using their chipsets, the more often they get to sell people an entire new board when they only need one small part.

    On-board video has been a disaster since day 1, for both PC and Mac users. What seems "high end" when a machine is new turns into "mediocre" within a year or two. Then come all the conflicts trying to get the on-board video disabled when you add a new, add-in video card. (I'm sure many long-time Mac users can remember the dislike for the "Performa" towers like the 6400/6500, largely due to the on-board video only allowing up to 2MB of video RAM.)

    Integrated NICs may work fine when they work, but again - I've seen many a blown NIC card due to power surges/spikes. I'd rather swap a card and have a fully functional machine again than have a dead port permanently soldered onto the back of my computer....

  27. security by spoonist · · Score: 3, Interesting

    I, for one, do not like this trend of integrating wireless into everything.

    As a security conscious individual, I want to be able to physically choose whether or not I want wireless when I want wireless.

    I like to be able to physically pull out the wireless card in my laptop because then I know I can't be h4x0r3d via my WLAN card.

    Fine. Call me paranoid. I don't mind.

    (Yeah, I know they said the RF part would still be an add-on... I'm just talking in general that I want add-ons and not fully integrated wireless stuff that I can't pull out without desoldering chips.)

  28. Dear God, would you people please think? by gmhowell · · Score: 3, Interesting

    Would you people please think before posting?

    This includes a chipset, not a radio. Therefore, it won't be sending out your world control schemes to everyone in existence. Yes, Intel will at some point in the nebulous future include a radio. As will many manufacturers. At that point, we go to the next paragraph:

    Every integrated soundcard/videocard/ethernet controller/serial port/etc. I've ever seen has a setting in the BIOS. If you don't want the location of your laser embedded sharks known to the black helicopter people, switch it off.

    Finally, when is the last time your built in ethernet card just randomly spewed data out the port to the CIA? Oh, last week? Then you have more problems than just a wireless AP built into your motherboard.

    --
    Jesus was all right but his disciples were thick and ordinary. -John Lennon
  29. Re:Hospitals by Jarrik · · Score: 3, Interesting

    I work in a fairly large hospital (200+ beds) and I can tell you from alot of experance that this is no longer the case for hospitals with modern equipment. The hospital I work for just had Nextel come in and install signal multiple repeaters on every floor and building of our entire campus and put a cell tower on top of our main building that houses all patent rooms / OR's with no problems what so ever. There are certain areas where we are advised not to have our Cell phones on but that is due more to respect of patent famlies and not becuase of any interferance issues.

  30. but... by sbma44 · · Score: 2, Informative
    you probably *could* have swapped in an ide controller board and just turned off the faulty integrated controller in the bios.

    I agree that we should not sacrifice modularity for all-in-one disposability, but for all the applications you list (IDE, NIC, video) you can put in a modular card and override the integrated stuff. Personally, I think ubiquitous integrated mobo NICs are one of the handiest hardware improvements of the last five years.

  31. The real problem is routing by Baldrson · · Score: 2, Informative
    Having set up a municipal mesh network it is pretty obvious the real problem to solve for WiFi isn't higher levels of integration so much as routing.

    Internet protocols were designed around wires and it shows when you go to wifi meshes. Meshes are critical due to the fact that meshes scale. If you are going to have a wifi node in every consumer device, as seems potentially viable, then you need to continually discover new routes and do so on nearly every packet. Route-flap is what you get, even with damping protocols, with current internet standards. You can end up waiting minutes for a route to stablize.

    Here's an algorithm for a mesh node that seems to work simply:

    Just keep a table of destination IP addresses in memory with a counter that decays exponentially with time.

    When the counter decays below some threshold, clip its IP address from the list. An IP address with no counter is considered to have a value of 0.

    Every time a packet acknowledgement comes through for a given destination IP address, add one to the counter for that IP address.

    Whenever a packet (not already awaiting acknowledgement) is 'heard' destined for an IP address, queue it for rebroadcast according to a priority established by the IP address's counter.

    Let packets that fall off the end of the queue due to low priority do so without further consideration.

    More complex algorithms are required for transmission power optimization, but even this simple algorithm shows how far off-base current internet protocols are for wifi.

    1. Re:The real problem is routing by sploxx · · Score: 2, Informative

      In IPV4, you can't allocate space for WiFi meshes anymore. ACK.

      But this is not true for IPV6. In IPV6 you can allocate big classless spaces and use these for your mesh networks. A 10 million city with one IP per PC, washing mashine, toaster... no problem.

      What you describe at the most important problem in todays IP world is not the basic protocol, i.e. IP. You are criticizing the routing protocols, they do not match well with fluctuating mesh networks. Indeed. They are designed for wired networks. But most of the internet traffic still runs over wires, and this will probably stay so even with competion from wifi. You can't really put all that bandwidth through wifi.

      There are solutions that work on top of IP, in a very clean way, and these solutions work well in the lab. In the field, there are not yet enough studies which support this research.
      The proposed new network architectures consist mainly of new routing protocols adapted to mesh networks. I see no problem with IP as the underlying protocol. It works, it's relatively reliable and it's already there. And the research groups (Are you professionally working on this topic? - I'm not. But this doesn't invalidate my arguments, I think...) are using IP on the lower level...

      For more information, try google for "manet". There is already much information out there.

      I can't really talk about you routing algorithm because I haven't seen it in simulation or reality. To state that a particular routing algorithm is good by just drawing it onto paper is a good start, but not a proof of concept. Maybe it works, maybe not. But try it out and build devices which connect to a mesh network (Mesh networking is one of the things I consider to play around with next in my freetime) and show the world that it works. But I'm sure that the underlying packets can IPV6 instead of your custom protocol without any functional penalty.

    2. Re:The real problem is routing by Baldrson · · Score: 2, Interesting
      But this is not true for IPV6. In IPV6 you can allocate big classless spaces and use these for your mesh networks. A 10 million city with one IP per PC, washing mashine, toaster... no problem.

      Well, yeah, and I did try to get Jon Anderson to include IPv6 in his first major release of LocustWorld's MeshAP for precisely that reason. To no avail however. He thinks the ability to set up private subnets, local to the mesh, is adequate to the problem at hand. He may be right if the problem at hand is simply getting out to a lot of MANs rapidly.

      As for my professional involvement, I am responsible for deployment and mantainence of a mesh network in a small metro area so yes to that extent I'm professionally involved. I'll probably do a computer simulation of the described algorithm before building any actual devices but yes testing of these ideas is important.

  32. Mobo features by t0ny · · Score: 2, Informative

    It will probably be easy for mobo makers to make use of this as an optional feature, much like SATA, USB 2.0, etc. Then you just need to have an antenna lead from the mobo, and enable it in the BIOS: it will work just like parallel/serial ports.

    --

    Manipulate the moderator system! Mod someone as "overrated" today.

  33. Re:Unintentional network compromises through bridg by drinkypoo · · Score: 2, Interesting
    The network name will probably be set to 'intel' and no encryption key set, just like linksys hardware comes up as 'linksys' with no WEP key. Windows configures new interfaces for DHCP by default. Hence if someone sat outside your office with a WAP (or a pc with a wifi nic acting like a WAP) providing dns service, with the network name set to 'intel' (in my example, anyway) then they could assign IPs to your machines and begin hacking.

    The solution is simple, however; Don't buy the systems with wifi in them, or be sure to configure them with a software image, unattended install, or similar feature to configure the card out of the box, or disable it.

    --
    "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
  34. Re:Wintel are both right. by bhtooefr · · Score: 2, Informative

    http://slashdot.org/articles/99/11/25/094246.shtml

    Actually, they already tried that. However, their Israel design team had it almost finished when they pulled the plug, and they were PISSED. Ah, well, they were happy when they got the Pentium M project - what the P4 shoulda been!

    BTW, have you played with a Cyrix MediaGX? If so, then you should be modded funny. If not, they need to make a mod for stupid - Cyrix's implementation SUCKED - 44-50MHz CPU speed was lost, and the video was worse than i810.

  35. Why is this on Slashdot? by jwr · · Score: 2, Interesting

    Why do I read about this on Slashdot, instead of an article about Intel declining to provide drivers or specifications for their wireless part of Centrino?