MAC address should be enough - TCP-IP

This is a discussion on MAC address should be enough - TCP-IP ; Every NIC has a unique address, which means that every networked device is unique. OUI is the gateway and end of the story, right?...

+ Reply to Thread
Results 1 to 16 of 16

Thread: MAC address should be enough

  1. MAC address should be enough

    Every NIC has a unique address, which means that every networked device is
    unique.
    OUI is the gateway and end of the story, right?



  2. Re: MAC address should be enough

    eager wrote:

    > Every NIC has a unique address, which means that every networked device is
    > unique.


    MAC addresses can very easily be changed and spoofed. MAC also only apply to
    ethernet-like networks. Other technologies use different types of
    addresses.

    > OUI is the gateway and end of the story, right?


    What do you mean? OUI is the name of the first three octets of a MAC
    address, which identify the manufacturer.


  3. Re: MAC address should be enough

    "eager" writes:

    > Every NIC has a unique address, which means that every networked device is
    > unique.
    > OUI is the gateway and end of the story, right?


    If you assume the MAC addresses are random, then how can a route be
    found for that device? It would require that EVERY system on the
    Internet know the MAC address for every other system, and the route as well.

    Let me make an understatement - that would be hard.


    --
    Posted via a free Usenet account from http://www.teranews.com


  4. Re: MAC address should be enough

    In article <6E2bj.1196$vd4.385@pd7urf1no>, eager wrote:
    >Every NIC has a unique address, which means that every networked device is
    >unique.
    >OUI is the gateway and end of the story, right?


    MAC addresses are only required to be unique for each LAN, not
    globally.

    Sometimes manufacturers make a mistake and duplicate a range of MAC
    addresses. It doesn't matter, unless someone happens by chance to
    receive a device that duplicates a MAC they already have. Sometimes
    manufacturers make a mistake and make short runs of devices with
    duplicate MAC addresses; it doesn't matter, unless someone happens
    to order several of the same device and gets delivery of more than
    one from the block that has duplicates.

    There are some devices for which the manufacturer programs the devices
    with the same MAC all the time, so that the client devices can
    locate the master device without needing an IP address.

    There are some cases involving redundancy in which several NICs are
    programmed with the same MAC address so that all of the NICs
    involved receive copies of the incoming packets.

  5. Re: MAC address should be enough


    "Walter Roberson" wrote in message
    news:Ogbbj.3724$DP1.748@pd7urf2no...
    > In article <6E2bj.1196$vd4.385@pd7urf1no>, eager wrote:
    >>Every NIC has a unique address, which means that every networked device is
    >>unique.
    >>OUI is the gateway and end of the story, right?

    >
    > MAC addresses are only required to be unique for each LAN, not
    > globally.
    >
    > Sometimes manufacturers make a mistake and duplicate a range of MAC
    > addresses. It doesn't matter, unless someone happens by chance to
    > receive a device that duplicates a MAC they already have. Sometimes
    > manufacturers make a mistake and make short runs of devices with
    > duplicate MAC addresses; it doesn't matter, unless someone happens
    > to order several of the same device and gets delivery of more than
    > one from the block that has duplicates.


    Why should the rest of world pay for manufacturers mistakes and go through
    so much trouble.
    Instead, a centralized organization such as IANA can take control over this,
    and every manufacture should register the MAC address for their cards the
    same as the case with registering domain names...
    The customers get a license of registration for the NICs they purchase and
    they can transfer the license if they need to replace a faulty NIC.

    if a 48 bit address is not enough, there can decide in a 128 bit address for
    that matter ....



  6. Re: MAC address should be enough


    "Bruce Barnett" wrote in message
    news:yek4peag1be.fsf@mail.grymoire.com...
    > "eager" writes:
    >
    >> Every NIC has a unique address, which means that every networked device
    >> is
    >> unique.
    >> OUI is the gateway and end of the story, right?

    >
    > If you assume the MAC addresses are random, then how can a route be
    > found for that device?


    The OUIs won't be random ....

    It would require that EVERY system on the
    > Internet know the MAC address for every other system, and the route as
    > well.


    That's the whole point! License registration, increasing security,
    simplicity ....

    Wait a minute, that's how it is right now, isn't it?
    Every system on the Internet should know the route to connect to another
    system.




  7. Re: MAC address should be enough


    "io" wrote in message news:fkimd8$14l$1@aioe.org...
    > eager wrote:
    >
    >> Every NIC has a unique address, which means that every networked device
    >> is
    >> unique.

    >
    > MAC addresses can very easily be changed and spoofed. MAC also only apply
    > to
    > ethernet-like networks. Other technologies use different types of
    > addresses.



    Make them 128 bit and registered.


    >
    >> OUI is the gateway and end of the story, right?

    >
    > What do you mean? OUI is the name of the first three octets of a MAC
    > address, which identify the manufacturer.


    use the OUI for routing.



  8. Re: MAC address should be enough

    "eager" wrote in message
    news:Apdbj.4435$DP1.850@pd7urf2no...
    >
    > "Bruce Barnett" wrote in message
    > news:yek4peag1be.fsf@mail.grymoire.com...
    > > "eager" writes:
    > >
    > >> Every NIC has a unique address, which means that every networked device
    > >> is
    > >> unique.
    > >> OUI is the gateway and end of the story, right?

    > >
    > > If you assume the MAC addresses are random, then how can a route be
    > > found for that device?

    >
    > The OUIs won't be random ....
    >
    > It would require that EVERY system on the
    > > Internet know the MAC address for every other system, and the route as
    > > well.

    >
    > That's the whole point! License registration, increasing security,
    > simplicity ....


    No - the OUI belongs to a manufacturer - that kit can be sprayed across the
    world, and every manufacturer can do what they like with the 24 bits they
    define as long as they follow some rules.

    OUI is more of a serial number than a global address - with the added
    complication that some devices get lots of them (a big switch might have a
    pool of 1000 or more MACs), and you can make a MAC yourself up using the
    "local address" option.

    FWIW IPv6 is the next addressing scheme designed to be "bigger" for the
    internet with 128 bit addresses- but it isnt really mainstream even now,
    approx 10 years after it got defined.
    >
    > Wait a minute, that's how it is right now, isn't it?
    > Every system on the Internet should know the route to connect to another
    > system.


    No - IP is deliberately designed so that is not needed.

    The IP (and the internet as the biggest deployed IP network) uses
    hierarchical addressing to make it feasible to approximate to "knowing the
    address of every other interface".

    So sending a packet from Europe to somewhere in USA it goes "thataway" to a
    transatlantic link.

    Exactly where in USA and which way to send it once it gets to the US E Coast
    is a problem only for the router it hits when it gets there (and on and on).

    Current backbone routing tables even with all this hierarchy is around 1M
    routes, and only high end core routers have to deal with that - edge boxes
    like a SOHO router between you and your ISP only need to know "local"
    addresses and "other", with increasing scale as the routers have to make
    more complex choices.

    But there are 1000M+ devices active in the Internet, and using unique non
    hierarchical addresses would be unmanageable, and v difficult to make work.

    Finally - the existing system has the built in inertia of any installed
    working system.

    In a sense it doesnt matter whether your scheme is "better" in some way, as
    you would have to get everyone involved to agree to change, and you would
    need a massive set of benefits for everyone invovled to agree to the hassle
    changing would cause.....
    >

    --
    Regards

    stephen_hope@xyzworld.com - replace xyz with ntl



  9. Re: MAC address should be enough


    "stephen" wrote in message
    news:1ifbj.11156$h35.2952@newsfe2-gui.ntli.net...
    > "eager" wrote in message
    > news:Apdbj.4435$DP1.850@pd7urf2no...
    >>
    >> "Bruce Barnett" wrote in
    >> message
    >> news:yek4peag1be.fsf@mail.grymoire.com...
    >> > "eager" writes:
    >> >
    >> >> Every NIC has a unique address, which means that every networked
    >> >> device
    >> >> is
    >> >> unique.
    >> >> OUI is the gateway and end of the story, right?
    >> >
    >> > If you assume the MAC addresses are random, then how can a route be
    >> > found for that device?

    >>
    >> The OUIs won't be random ....
    >>
    >> It would require that EVERY system on the
    >> > Internet know the MAC address for every other system, and the route as
    >> > well.

    >>
    >> That's the whole point! License registration, increasing security,
    >> simplicity ....

    >
    > No - the OUI belongs to a manufacturer - that kit can be sprayed across
    > the
    > world, and every manufacturer can do what they like with the 24 bits they
    > define as long as they follow some rules.
    >
    > OUI is more of a serial number than a global address - with the added
    > complication that some devices get lots of them (a big switch might have a
    > pool of 1000 or more MACs), and you can make a MAC yourself up using the
    > "local address" option.
    >
    > FWIW IPv6 is the next addressing scheme designed to be "bigger" for the
    > internet with 128 bit addresses- but it isnt really mainstream even now,
    > approx 10 years after it got defined.
    >>
    >> Wait a minute, that's how it is right now, isn't it?
    >> Every system on the Internet should know the route to connect to another
    >> system.

    >
    > No - IP is deliberately designed so that is not needed.
    >
    > The IP (and the internet as the biggest deployed IP network) uses
    > hierarchical addressing to make it feasible to approximate to "knowing the
    > address of every other interface".
    >
    > So sending a packet from Europe to somewhere in USA it goes "thataway" to
    > a
    > transatlantic link.
    >
    > Exactly where in USA and which way to send it once it gets to the US E
    > Coast
    > is a problem only for the router it hits when it gets there (and on and
    > on).
    >
    > Current backbone routing tables even with all this hierarchy is around 1M
    > routes, and only high end core routers have to deal with that - edge boxes
    > like a SOHO router between you and your ISP only need to know "local"
    > addresses and "other", with increasing scale as the routers have to make
    > more complex choices.
    >
    > But there are 1000M+ devices active in the Internet, and using unique non
    > hierarchical addresses would be unmanageable, and v difficult to make
    > work.
    >
    > Finally - the existing system has the built in inertia of any installed
    > working system.
    >
    > In a sense it doesnt matter whether your scheme is "better" in some way,
    > as
    > you would have to get everyone involved to agree to change, and you would
    > need a massive set of benefits for everyone invovled to agree to the
    > hassle
    > changing would cause.....
    >>

    > --


    These rules can be the same as the country /area codes for OUI and the same
    as the phone numbers for the rest of the MAC address in the
    telecommunication system!
    Even if ISPs can map the serial numbers / MAC addresses of NICs into the
    telephone numbers, it will be easier especially to find out where the
    message/spam/attack/ etc. is coming from.




  10. Re: MAC address should be enough

    On Dec 22, 4:36*pm, "eager" wrote:

    > These rules can be the same as the country /area codes for OUI and the same
    > as the phone numbers for the rest of the MAC address in the
    > telecommunication system!
    > Even if ISPs can map the serial numbers / MAC addresses of NICs into the
    > telephone numbers, it will be easier especially to find out where the
    > message/spam/attack/ etc. is coming from.- Hide quoted text -


    I think you have understood the problem. There is nothing inherent in
    the MAC address to indicate where the packet has to be routed. The OUI
    doesn't help, since all it does is identify the manufacturer of the
    card. And, as others have said, the MAC address can be changed. Some
    protocols always did that.

    However, you might be happy to know that IPv6 does use the
    (theoretically unique) MAC address within the 128-bit IP address, as
    one of the addressing options. See RFC 2460 for how that works.

    Alberto

  11. Re: MAC address should be enough

    "eager" wrote in message
    news:xrfbj.2891$vd4.2815@pd7urf1no...
    >
    > "stephen" wrote in message
    > news:1ifbj.11156$h35.2952@newsfe2-gui.ntli.net...
    > > "eager" wrote in message
    > > news:Apdbj.4435$DP1.850@pd7urf2no...
    > >>
    > >> "Bruce Barnett" wrote in
    > >> message
    > >> news:yek4peag1be.fsf@mail.grymoire.com...
    > >> > "eager" writes:
    > >> >
    > >> >> Every NIC has a unique address, which means that every networked
    > >> >> device
    > >> >> is
    > >> >> unique.
    > >> >> OUI is the gateway and end of the story, right?
    > >> >
    > >> > If you assume the MAC addresses are random, then how can a route be
    > >> > found for that device?
    > >>
    > >> The OUIs won't be random ....
    > >>
    > >> It would require that EVERY system on the
    > >> > Internet know the MAC address for every other system, and the route

    as
    > >> > well.
    > >>
    > >> That's the whole point! License registration, increasing security,
    > >> simplicity ....

    > >
    > > No - the OUI belongs to a manufacturer - that kit can be sprayed across
    > > the
    > > world, and every manufacturer can do what they like with the 24 bits

    they
    > > define as long as they follow some rules.
    > >
    > > OUI is more of a serial number than a global address - with the added
    > > complication that some devices get lots of them (a big switch might have

    a
    > > pool of 1000 or more MACs), and you can make a MAC yourself up using the
    > > "local address" option.
    > >
    > > FWIW IPv6 is the next addressing scheme designed to be "bigger" for the
    > > internet with 128 bit addresses- but it isnt really mainstream even now,
    > > approx 10 years after it got defined.
    > >>
    > >> Wait a minute, that's how it is right now, isn't it?
    > >> Every system on the Internet should know the route to connect to

    another
    > >> system.

    > >
    > > No - IP is deliberately designed so that is not needed.
    > >
    > > The IP (and the internet as the biggest deployed IP network) uses
    > > hierarchical addressing to make it feasible to approximate to "knowing

    the
    > > address of every other interface".
    > >
    > > So sending a packet from Europe to somewhere in USA it goes "thataway"

    to
    > > a
    > > transatlantic link.
    > >
    > > Exactly where in USA and which way to send it once it gets to the US E
    > > Coast
    > > is a problem only for the router it hits when it gets there (and on and
    > > on).
    > >
    > > Current backbone routing tables even with all this hierarchy is around

    1M
    > > routes, and only high end core routers have to deal with that - edge

    boxes
    > > like a SOHO router between you and your ISP only need to know "local"
    > > addresses and "other", with increasing scale as the routers have to make
    > > more complex choices.
    > >
    > > But there are 1000M+ devices active in the Internet, and using unique

    non
    > > hierarchical addresses would be unmanageable, and v difficult to make
    > > work.
    > >
    > > Finally - the existing system has the built in inertia of any installed
    > > working system.
    > >
    > > In a sense it doesnt matter whether your scheme is "better" in some way,
    > > as
    > > you would have to get everyone involved to agree to change, and you

    would
    > > need a massive set of benefits for everyone invovled to agree to the
    > > hassle
    > > changing would cause.....
    > >>

    > > --

    >
    > These rules can be the same as the country /area codes for OUI and the

    same
    > as the phone numbers for the rest of the MAC address in the
    > telecommunication system!
    > Even if ISPs can map the serial numbers / MAC addresses of NICs into the
    > telephone numbers, it will be easier especially to find out where the
    > message/spam/attack/ etc. is coming from.


    So "all" you are trying to do is work out where something came from?

    the current system does that anyway - it ties an IP source to an email
    source, the IP belongs to an address block, and that block is specific to an
    ISP, and they should be able to work out which of their users IP addresses
    it came from.

    If your ISP runs some sort of spam management, then if you look in the
    headers of an email you will find the info you want.

    Or analyse what is there with a tool such as Sam Spade - it will look up the
    set of headers.

    But - if that end point address is a zombie PC, then the PC isnt where the
    content of the email came from, so that doesnt help as the owner of the PC
    didnt originate the email......

    finally - i fail to see how your new addressing will help - tieing an email
    to a new kind of address is just the same as tieing it to a source IP
    address as you can do now.

    It doesnt "fix" the spam problem, which is all about a spam source using
    other people machines & resources to send the emailed garbage.

    If you care about spam, then report the ones you get to the source ISPs, the
    black lists and to others referenced, such as the operator of the server /
    address / DNS for phishing servers, affected banks et al.
    >

    --
    Regards

    stephen_hope@xyzworld.com - replace xyz with ntl



  12. Re: MAC address should be enough

    "eager" writes:

    > Wait a minute, that's how it is right now, isn't it?
    > Every system on the Internet should know the route to connect to another
    > system.


    Not at all. My network knows two routes.

    Everything on my LAN is delivered directly to the device
    Everything else in the world goes to my default router.

    More sophisticated networks may have more routes. And some backbone
    routers have a harder job. But they aren't $100 devices either.

    --
    Posted via a free Usenet account from http://www.teranews.com


  13. Re: MAC address should be enough

    eager wrote:

    >> MAC addresses can very easily be changed and spoofed. MAC also only apply
    >> to ethernet-like networks. Other technologies use different types of
    >> addresses.

    >
    > Make them 128 bit and registered.


    That might solve some problems, but not the main one (see below). Moreover,
    lots of network technologies exist that do not use MAC addresses for
    addressing (eg, frame relay to name one).

    >>> OUI is the gateway and end of the story, right?

    >>
    >> What do you mean? OUI is the name of the first three octets of a MAC
    >> address, which identify the manufacturer.

    >
    > use the OUI for routing.


    To avoid having each device know the route to every other device in the
    network, addresses must have some kind of structure embedded, so that
    hierarchical addrssing and summarization are possible. For IP, the
    structure is defined by the subnet mask, which permits address aggregation
    and summarization (meaning that increasingly shorter subnet masks can be
    used as routes are propagated upstream, to reduce the size of the routing
    tables in core routers - ok, I know this is quite oversimplified but I hope
    you get the idea). This works based on the assumption that addresses are
    distributed in such a way that addresses that are numerically "close" also
    are geographically close. This is not exactly what's happening in the
    Internet of today, but the current distribution is still better than having
    no hierarchy at all.

    OUIs do not have the requisites above, because network cards produced by a
    given manufacturer (thus with the same OUI) can be located everywhere
    around the world. If the OUI were used for routing, *each* network router
    would have to know exactly and at every time where *each* network device in
    the world is. No need to say that their routing tables would not be very
    stable, since at each time there would be lots of devices changing their
    location in the network (eg, notebooks or wireless devices). The traffic
    produced by all those routing updates would be overwhelming (remember that
    *each* single location change would have to be propagated to *each* single
    router), as would be the load placed on the routers' processors to process
    those updates. And I'm deliberately ignoring the amount of memory that
    would be needed (again, in *each* router) to keep routing tables of that
    size in RAM (think of, say, 16 bytes for each entry, for maybe a billion
    devices in the network...more or less about 14GB of RAM per router, but
    this is surely underestimated).


  14. Re: MAC address should be enough


    "pk" wrote in message news:fklsi5$d8r$1@aioe.org...
    > eager wrote:
    >
    >>> MAC addresses can very easily be changed and spoofed. MAC also only
    >>> apply
    >>> to ethernet-like networks. Other technologies use different types of
    >>> addresses.

    >>
    >> Make them 128 bit and registered.

    >
    > That might solve some problems, but not the main one (see below).
    > Moreover,
    > lots of network technologies exist that do not use MAC addresses for
    > addressing (eg, frame relay to name one).


    Frame relay is a layer 2 protocol!

    >
    >>>> OUI is the gateway and end of the story, right?
    >>>
    >>> What do you mean? OUI is the name of the first three octets of a MAC
    >>> address, which identify the manufacturer.

    >>
    >> use the OUI for routing.

    >
    > To avoid having each device know the route to every other device in the
    > network, addresses must have some kind of structure embedded, so that
    > hierarchical addrssing and summarization are possible. For IP, the
    > structure is defined by the subnet mask, which permits address aggregation
    > and summarization (meaning that increasingly shorter subnet masks can be
    > used as routes are propagated upstream, to reduce the size of the routing
    > tables in core routers - ok, I know this is quite oversimplified but I
    > hope
    > you get the idea). This works based on the assumption that addresses are
    > distributed in such a way that addresses that are numerically "close" also
    > are geographically close. This is not exactly what's happening in the
    > Internet of today, but the current distribution is still better than
    > having
    > no hierarchy at all.
    >
    > OUIs do not have the requisites above, because network cards produced by a
    > given manufacturer (thus with the same OUI) can be located everywhere
    > around the world. If the OUI were used for routing, *each* network router
    > would have to know exactly and at every time where *each* network device
    > in
    > the world is. No need to say that their routing tables would not be very
    > stable, since at each time there would be lots of devices changing their
    > location in the network (eg, notebooks or wireless devices). The traffic
    > produced by all those routing updates would be overwhelming (remember that
    > *each* single location change would have to be propagated to *each* single
    > router), as would be the load placed on the routers' processors to process
    > those updates. And I'm deliberately ignoring the amount of memory that
    > would be needed (again, in *each* router) to keep routing tables of that
    > size in RAM (think of, say, 16 bytes for each entry, for maybe a billion
    > devices in the network...more or less about 14GB of RAM per router, but
    > this is surely underestimated).
    >


    IPv6 explains all that .... but why using overheads ?

    i.e. DTE - DCE use only the first two layers ....



  15. Re: MAC address should be enough

    eager wrote:

    >> lots of network technologies exist that do not use MAC addresses for
    >> addressing (eg, frame relay to name one).

    >
    > Frame relay is a layer 2 protocol!


    As is ethernet, where MAC addresses are used.

    > IPv6 explains all that .... but why using overheads ?
    >
    > i.e. DTE - DCE use only the first two layers ....


    You can't know for sure in advance that every network connected to the
    Internet will use the dte-dce paradigm, nor you can force people to use a
    predetermined technology. You need a protocol that abstracts from the layer
    2 details and that can run virtually on every kind of layer 2 technology.
    This is what IP does, and this is certainly a lot more flexible than
    requiring that each network use a given layer 2 technology (in addition to
    all other benefits already mentioned).


  16. Re: MAC address should be enough

    On Dec 22, 1:36 pm, "eager" wrote:

    > These rules can be the same as the country /area codes for OUI and the same
    > as the phone numbers for the rest of the MAC address in the
    > telecommunication system!
    > Even if ISPs can map the serial numbers / MAC addresses of NICs into the
    > telephone numbers, it will be easier especially to find out where the
    > message/spam/attack/ etc. is coming from.


    Why not just have every spammer put has name, address and telephone
    number in every spam he sends. That would save having to go to the
    ISP's map.

    DS

+ Reply to Thread