Seems to me there maybe some confusion here between the workings of ethernet, netbios and tcp/ip - netbios functions in the absence of tcp/ip so the concept of a netbios broadcast looking for a reply containing an ip address seems more than a little incorrect.
Really? First of all I tried to simplify my explanation and make it as short as possible in order not to overwhelm a reader with unnecessary details. But I still do not see where is the confusion?
NetBIOS works on 5th (session) layer of OSI model, while UDP protocol is on Layer 4 (transport) layer and IP is on Layer 3 (network) layer.
Here is the full model of communications according to OSI:
Layer 7. Application. Applications like Browser work on this layer using, SMB or HTTP protocols
Layer 6. Presentation. Determines formatting of data for Layer 7 (ASCII, JPEG etc.)
Layer 5. Session. Provides protocol of signalling and data exchenge between applications - NetBIOS for example with its set of queries and responses.
Layer 4. Transport. Provides Transport protocol for delivering data obtained from layers above. It may be UDP, TCP, ESP. AH etc for TCP/IP stack, may be NetBEUI, may be SPX for IPX/SPX stack etc. Layers above do not care much about type of protocol used at this level.
Layer 3. Network. Provides network protocol which is responsible for delivering data formed at the above level. For TCP/IP stack it is IP protocol, for IPX/SPX it is IPX, for NteBEUI ... it is NetBEUI itself as it is very simple and covers both levels at the same time. There again may be other protocols as well. Layers 5 -7 do not care which ones. Routers are called Layer 3 devices because they operate at this layer (and below.)
Level 2. Data link. It defines protocol of data transmission between 2 nodes. This may be Ethernet, Token Ring, PPP, Serial, ATM, Frame Relay etc. Switches are L2 devices because they operate at this leayer and below. They have no idea about IP, IPX, Apple Talk, Powerline protocols etc. They know about MAC addresses only. BTW MAC address is used only by some of L2 protocols. Serial cards do not have them for example. This layer does not care about Layer 1.
Level 1. Physical. This layer provides physical media for layer above - twisted copper pair, electrical wiring (for powerline adapters), optical fiber, Wi-Fi, CDMA, 3G, GPRS, GSM, etc.
When computer needs to transfer data to another host, data passes through all the layers top to bottom, and then receiver pulls up all this data from bottom to top to target application. Task of each layer is to wrap data from previous one to format appropriate to the same layer on receiver host and pass it to layer below. It does not care about the way that layer will process it. Very good analogy with postal service. We do not care what type of vehicles post uses to deliver our mail to recipient. We write a mail in a format (language) appropriate for receiver and put it in the envelope (transport layer), write source and destination address (network-routing level) and drop into PO box (data link layer - PO logistics). Physical layer (car, airplane) transmits it point to point where everything goes in back direction.
Again, NetBios or browser do not care about type cable or network protocol used as long as it is able to accomplish the task (like broadcasting).
In all good days of Windows for workgroups and Novell NetWare we had NetBEUI drivers installed for communications between computers in a workgroup on the local LAN (as NetBEUI is a non-routable protocol), IPX/SPX driver to access folders and printers on NetWare servers and TCP/IP to browse internet. And the same explorer application could use all protocols at the same time without problems.
Nowadays 99% of LAN use TCP/IP stack as Layer3/4, Ethernet as Layer 2, for Layer 1 - UTP for wired and Wi-Fi for Wireless. I also use powerline :-).
So when we a talking about NetBIOS we do not care about NetBIOS over NetBEUI (since windows XP came) or NetBIOS over SPX/IPX. We talking about NetBIOS over TCP/IP aka NBT.
Now if we trace how communication takes place we will see approximately following:
Browser application gets request to access \\DNS-323\Volume_1 share.
To resolve DNS-323 name it goes into netbios cache of resolved names and does not find it. Let's assume that LMHOSTS file is empty and WINS server is not configured at all.
Then as a last resort (before switching to DNS) it sends broadcast (using UDP/137) to local network asking "Who's name DNS-323?".
If DNS-323 is online it responds with unicast destined to the PC with response which contains its IP address.
After that PC places new entry into netbios cache with name-IP mapping (just like DNS resolution works)
Now as PC has IP destination IP address it needs to know MAC address of NIC card to which data should be tranferred. To do this it checks obtained IP with its own IP and address mask to figure out if that IP is in the same subnet.
If it is (which is very likely as broadcasts by default never forwarded by routers) then it checks its dynamic ARP table in hope to find IP-to-MAC mapping for DNS-323 IP address.
If that IP address happened to be on another network, PC will look for MAC address of Default Gateway (or gateway which was explicitly configured for that network in PC routing table using ROUTE PRINT command for example).
In either cases PC first checks its ARP table for existing entries. If there is no IP it is looking for, it performs ARP request (broadcast again) to the network asking "Who has X.X.X.X? Answer X.X.X.Y."
When DNS receives ARP request with its IP it sends back to the requester unicast response with its MAC address.
PC receives the response and adds IP-MAC mapping into its ARP table. Now application (browser) is ready to send a request to the NAS.
Formed request (say "obtain directory list of \Volume_1") is passed to netbios which establishes session with NAS and transmits datagrams to it.
NetBios requests in case of using TCP/IP are encapsulated into UDP datagrams with destination port 139 for session messages and port 138 for data itself.
Network layer (3) forms IP packets by adding IP header with source and destination IP, protocol type (UDP) and other fields.
Data link layer forms ethernet frames by addind ethernet header with source and destination MAC addresses taken form ARP table, type of protocol inside (IP in this case), append checksum and forwards it to Layer 1.
On layer 1 formed frame is finally transmitted to the recipient NIC over physical media.
I think I omit details of CDMA/CD or CDMA/CA and other L1 protocols operation
.
Receiver (DNS-323) receives transmission and "pulls" it from L1 to L7 to obtain data for application to process and form a response.
So, where is a confusion?
I don't know if you're aware of this, but not only are ip addresses NOT required for local area network communications, they are NOT used, even when tcp/ip is the only network protocol in play. In local area network communications it is the MAC addresses that are used to identify the source & destination hosts rather than the ip addresses.
That's a news for me. Having passed a bunch of cisco exams I haven't heard that if all the machines on the same LAN, IP or other L3 addresses are not necessary and may be omitted, and Layer 5 can pass data directly to Layer 2 for encapsulation and transmission
Then why Microsoft bothered at all to create NetBEUI - simple small non-routable L3/L4 protocol aimed to work inside single LAN with source and destination netbios names as source and destination addresses for initial connection setup?
Did they also confuse OSI layers?
Could you explain please?
Author
Topic: ip address (Read 49268 times)