brouter - bridging non routable (layer 3?!) addresses - terminology question

A brouter is said to bridge non routable prtocols. But aren't non routable protocols still Layer 3? I guess it's not routing 'cos it doesn't involve a routing table. Functionally I suppose it's bridging. So I guess it's bridging at layer 3?

so a brouter does either bridging at layer 3 or routing at layer 3 - depending on the protocol.

I had never heard of bridging at layer 3.

am I right?

I know brouter is considered by some to be not such a serious term. But it does bring up that OSI terminology question.

Reply to
q_q_anonymous
Loading thread data ...

Well, Appletalk has a similar layer structure to TCP/IP, including the ability to be routed. Some IP routers will bridge Appletalk, such that the IP network is divided into subnets, but the Appletalk network is one large net.

There are also non-routable protocols, which pretty much don't have a layer 3, so bridging is the only way.

-- glen

Reply to
glen herrmannsfeldt

Not usually. For example, classic IPX does not involve IP addresses.

There are a number of short-range protocols with no IP addressing. For example, Spanning Tree, CDP, BPDUs; the Appletalk that the previous poster mentioned.

Reply to
Walter Roberson

I'm wondering if you just misunderstood his question,Or I misundertand your answer, since your answers are usually pretty dead on. You seem to be implying that a protocol is not layer 3 if it does not use IP addresses.

IPX uses no IP address, but it is still a layer 3 protocol. It does have IPX network and station addresses. Even tho the station address and the MAC address are one in the same, IPX is still a Layer 3 protocol.

And IPX *IS* a *routable* layer 3 protocol. IPX routers have been around for over 20 years now. In fact, any netware 2.x, 3.x or 4.x box that has NIC's attaching it to more than one network will act as an IPX router by defualt,a nd will send out IPX RIP packets once per minute. You have to specifically turn it off. And of course most major routers will route it as well. Bay/RS routers I know support it.

Just beasue something does not use IP addresses, or is non-routable, does not mean it is not layer 3. Spanning Tree and it's BPDU's, as well as GARP, etc, I would argue are not layer 3 but layer 2, since they carry info specifically for layer 2 use. Appletalk I would argue IS layer 3.

Reply to
snertking

Some are, some are not. I would not call Spanning Tree a layer 3 protocol. It is a protocol, and it is not routable (nor would one WANT to route it!) and it carries ethenet specific information of interest only to connected switches. A counter expample is Netbios/Netbui over erthernet: not routable either, but it IS a layer 3 protocol. DEC LAT is another good example of a non-routable layer 3 protocol.

terms are used in such a nebulous manner by the imps that work in marketing that I'd need to know exactly what device you are referring to to be able to answer. I have not seen the term used to describe any device in years. Generally what BROUTER means is that it routes the protocols it knows how to handle at layer 3, and bridges everything else at layer 2.

bridging is layer 2 by definition.

Right on which question?

Reply to
snertking

That sounds like Cisco's "IRB" feature, "Integrated Routing and Bridging", But I've never heard of Cisco using the term "brouter".

It also sounds like what the Bay Accelar 1100/1200 does when VLANs of different types are associated with a port: an incoming packet of a non-routable protocol is bridged (switched) to all other ports with the same VLAN number, but incoming routable packets are examined and switched if possible but routed to a different vlan if the destination IP is in the subnet associated with that VLAN number. I've never heard of Bay [Nortel] calling this a "brouter".

Reply to
Walter Roberson

That's arguable. NetBIOS/NetBEUI is more of an application-layer interface to the Data Link than an internetworking (Layer 3) protocol.

DEC LAT is a good example of an Application layer protocol (or arguably, a Transport (Layer 4) protocol). It provides terminal services analogous to Telnet, although it is restricted to a single Data Link due to the fact that it interfaces directly to the Ethernet and incorporates no internetworking (Layer 3) capability. It is most definitely not a Layer

3 protocol.
Reply to
Rich Seifert

Yeh, I've gotten so accustomed to thinking of "router" as meaning "IP router".

Reply to
Walter Roberson

Layer 3 because it's a logical address?

so if a device 'relays/forwards' using NETBIOS/netbeui over eth, is it bridging or routing?

A Brouter is said to bridge nonroutable protocols. That indicates that the word bridging is being used at Layer 3 Yet later, you say bridging is layer 2 only.

Generally what BROUTER means is that it routes the

but surely it's not bridging everything else at layer 2. NETBIOS you said is a layer 3 so if it deals with that one, it's bridging at layer

  1. I wonder if the term non routable isn't OSI. I guess it's only called non routable because it's flat. But maybe OSI considers it routable 'cos it's layer 3. (assuming layer 3 means logical)

in which case, saying 'bridging netbios' would mean ignoring the layer

3 netbios part, and bridging the - say - MAC addresses.

thanks

Reply to
q_q_anonymous

is there a layer 3 protocol with a flat namespace or is that a contradiction in terms?

(I seem to recalll that you prob know your OSI back to front 'cos I remember a post of yours that said the physical layer PDU (electricity i guess) is termed a symbol. that's a rare piece of knowledge I coudln't find anywhere else!)

Reply to
q_q_anonymous

When one says "bridging if it can't route," what that means is that whatever frames arrive with an unknown Layer 3 header, or no Layer 3 header at all, remain within the domain of the local Link Layer. In an IP example, that would mean they stay in the same IP subnet as the brouter port on which they arrived. They are switched through the brouter as if this were a bridge, using only the MAC address information, and not forwarded to the port(s) that lead beyond the IP subnet of the port on which these frames arrived.

Layer 3 means beyond the scope of the local Link Layer protocol. In an IP over Ethernet example, beyond the boundaries of that local Ethernet catenet, for example. Layer 3 addresses tend to be hierarchical, but then again, some addresses that are used as Layer 2 can also be (e.g. NSAPAs used with ATM). And I suppose one could have non-hierarchical Layer 3 addresses too, in principle. It would require complete routing tables at each router.

Bert

Reply to
Albert Manfredi

As Bert Manfredi noted in a related post:

That is correct. If there was a Layer 3 protocol with a flat address space (not "name" space--minor nit), then every router would need to know the relative location of every device in the internet. The effect would be to make the entire internetwork one big, flat space, which is exactly what Layer 2 bridges do (for catenets rather than internets).

And it is precisely this behavior that makes hierarchical Layer 3 routing attractive--it breaks up the huge flat space into smaller, administratively manageable spaces, and makes the routing table problem manageable as well.

Reply to
Rich Seifert

Thanks. Problem was with my understanding of OSI layers apparrently.

Got it now, I think.

Reply to
snertking

It helps if you think of the layers as approximations. (though 1-3 aren't so bad.)

There is a story about when the OSI layer model was being developed, the Europeans wanted six and US wanted 8 layers (or the other way around) and compromised on seven.

Another story is that the ATM 48 byte payload is a compromise between Europe's 32 and US 64 (or the other way around).

-- glen

Reply to
glen herrmannsfeldt

Albert Manfredi wrote

that sounds very odd.

If the brouter sends the packet back out the interface where it came, then preumably, it means it was a mistake for the comp to send it there in the first place with that interface's Dest MAC.

If the brouter knows the correct MAC then it must have MACs for that funny layer 3 protocol, at every interface, for each comp on that interface.

If the brouter has the correct MAC, it means the host is there on that subnetwork, and the packet was sent with the wrong MAC.

If it doesn't have the correct MAC, i guess it'd drop it.

this is rather inefficient. If what i've said is correct, the brouter has all MACs for all hosts on all interfaces at each interface (otherwise who is it going to send the packet back to? itself? the original dest mac on the packet was itself) So, if it has them all, and the packet is for a subnet local to the brouter (though not local to the host) then our brouter would be able to send it out the correct interface with comp's MAC. Yet you are saying it wouldn't do that, since I suppose that'd be considered routing. (though surely it shouldn't have gone to the brouter in the first place, the sending comp should've had the MAC)

minor additional point - if by the brouter bridging you mean sending back out the same interface , that is very unbridge like. Bridges send out a diff interface. They choose the correct interface.

thanks

Reply to
q_q_anonymous

Let's say the brouter is acting as a bridge between multiple ports belonging to one particular IP subnet, and then it has a single port to another IP subnet somewhere else. Call the distant IP net "Subnet B," and the local one "Subnet A."

Any frame arriving at one of the Subnet A ports addressed to a Subnet A destination will clearly not go to that Subnet B port. It will be either flooded to all Subnet A ports (e.g. an ARP broadcast), or it will be switched to a particular Subnet A port which has been "learned," or it could be multicast to several ports, including possibly the Subnet B port.

So what happens if a frame with unknown Network Layer address arrives at one of the Subnet A ports? It can't be sent to the Subnet B port, or the box wouldn't be much of a router. So I'm saying that frame with unknown Layer 3 address can be switched to other Subnet A ports of the brouter, just like any bridge would do, but not to the Subnet B port.

How does it "know the correct MAC"? Because presumably it learned the MAC address by inspecting frames sourced from a particular port. MAC address xyz lives at port #5, let's say, and this port #5 is one of the Subnet A ports.

Okay, so now take the legitimate case in which there are two identical MAC addresses xyz, one in Subnet A and one in that distant Subnet B. If a frame from one of the other Subnet A ports is addressed to MAC address xyz, but contains no Layer 3 header, how should the brouter handle that case?

I'm saying, the brouter simply switches the frame from one Subnet A port to port #5, which also belongs to Subnet A, but not to the Subnet B port.

Well ... There's typically no reason to send a layer 2 frame back to the same port from which it was received. Although this behavior is certainly possible for a router. Take the case of a router supporting two IP subnets on the same physical Ethernet. If a frame arrives addressed to the other IP subnet, the router will take in the frame and spit it right back out again on the same port on which it was received.

So this too could happen with this brouter, if two IP subnets happen to live off the same port of the box.

Bert

Reply to
Albert Manfredi

Ohhhh I didn't realise a Brouter had that.

ok!

but then how is that different from a router with a switch connected at each interface?

I see the logic is slightly different (no deciding if a protocol is known/routable or not) but the effect is the same.

( I've think of my 'home router' as being like a 2 port router with a switch at my side so everything hits the switch first. )

is there a name for that?

many thanks

Reply to
q_q_anonymous

For a net that is mostly IP, but not completely, if it receives any packets other than X'0800' or X'0806' they are passed to the bridge logic, otherwise to the IP routing logic.

It might be that there is enough logic at the MAC level to do that without storing all the X'0800' packets not addressed to that port, otherwise it has to store and discard them.

Some can router or bridge IP, router or bridge Appletalk, and bridge any other packet type that comes through.

-- glen

Reply to
glen herrmannsfeldt

In my view, functionally identical. But all in one box.

I've heard it called a "one-armed router." I don't know if there's a more legit-sounding techie name.

Bert

Reply to
Albert Manfredi

thanks, can I ask you where you got that as a definition of brouter?

or is there a brouter you used, do you know the model num?

Reply to
q_q_anonymous

Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.