Talk:Router (computing)/Archive 1

Picture needed
Could we have a GFDL-free picture of something like a big Juniper router for this article, please? -- The Anome 23:02, 15 May 2004 (UTC)
 * Agreed. Consumer-grade routers are piddly compared to REAL routers. --CCFreak2K 10:31, 5 December 2006 (UTC)

Hubs
You know what might be swell for the countless neophytes who with the advent of household broadband Internet access are now encountering switches, routers and hubs in their everyday lives and don't know what they are.. It might be nice to see a differentiation between a switch and a hub, which many would assume (but not altogether correctly) does the same thing that a switch does. Don't look to me to put the information down; I don't know it! :)Bstct 16:06, 24 December 2006 (UTC)

Well it would be more suitable to compare a switch to a hub in one of their articles as this is about routers. Plus, the home "Router" isn't a true router as it doesn't truly route any information but just shares the internet connection up. Cdscottie 3:28, 09 January 2007 (UTC)

The home Cable and DSL "Router" is a true router. Its "Operating Mode" defaults to "Gateway", making it an edge router or a gateway router, which serves DHCP to the connected computer(s), performs NAT on traffic routed to the Internet, etc. "Operating Mode" can also be changed to "Router", which makes it a "router" router, which i guess is a weak core router. -Whiner01 05:38, 22 April 2007 (UTC)

What, no Zyxel!?
Why isn't Zyxel on the list. —The preceding unsigned comment was added by Q9R (talk • contribs) 11:11, 7 May 2007 (UTC).

Get rid of those lists?
A large portion of the article is a list of vendors (or other suppliers). That doesn't seem to belong in an encyclopedic article. It doesn't seem to be blatant advertising, but it also doesn't add much of value. And even at its current excessive length, the list is unlikely to be at all complete.

How about getting rid of it? Paul Koning 16:45, 3 May 2007 (UTC)
 * The list is inappropriate for the article and a spam magnet. I support getting rid of it. Nposs 17:08, 3 May 2007 (UTC)


 * I concur. The list of manufacturers just invites kiddies to add the name of their favorite toy NAT gateway.  --Afed 17:23, 11 May 2007 (UTC)

Removed an edit that brought back a whole bunch of old stuff
I undid an edit by User:Giftlite. The edit summary didn't seem to match what changed -- it effectively undid many of the earlier changes today. Giftlite, if I misunderstood your intent, my apologies. Paul Koning 00:55, 12 May 2007 (UTC)


 * You're absolutely right. I'm glad you caught this error, and I've made the change I intended to make last week. Thanks again, Paul. —The preceding unsigned comment was added by Giftlite (talk • contribs) 15:12, 14 May 2007 (UTC).

Semi-Protection
Requested this page be permantly semi-protected. This article had an expert tag on it. Feel semi-protection will somewhat help mis-infomation being re-placed into this article. Article has incorrect data in it for over a year and talk page does not support opinions it is just incorrect info being entered.

--User: (talk • contribs • count) 23:54, 22 June 2007 (UTC)


 * Alas, that's not what page protection is about. It's more about volume.  I was surprised that my request to semi-protect Gerritsen Beach, Brooklyn was turned down on grounds that the volume of abusive edits this month didn't warrant it.  As it happened, my request attracted an admin who noted that the inserted garbage was defamatory of a named individual, so the page was protected on that basis. Jim.henderson 02:49, 23 June 2007 (UTC)

What does not belong here
Please do not include RFCs that deal with routing protocols here. Also do not include infomation about routing there has been discussion about all of this on the routing talk page. It seems best to keep this information seperate, so the average user is not overwhelmed with the data. If you include this data, your edits will be reverted. Please post your information to the proper articles. as it is taken a great deal of time to clean up this mess. --User: (talk • contribs • count) 02:44, 26 June 2007 (UTC)

Update
The article is a little misleading- it states that routing happens at layer 3 of the OSI reference model, but we are always using some implementation of that model which is usually TCP/IP.

--I disagree; the TCP/IP model is what they use in Universities because it is more abstract, so it is easier to teach. I can state as a fact that Cisco's Networking Academy was still focusing on the OSI model as of May 2003 because if you want an in-depth discussion or to get down to the "nitty gritty", OSI is what you should be using. For my CCNA exams we had to know the OSI model backwards and forewards and just recognize the differences between it and the TCP/IP model. If they were to use the TCP/IP model in the discussion, the distinction between a router and an "intelligent switch" (which is what you buy from someplace like BestBuy that claims to be a router which it is not) becomes much harder to explain. The article should, however, point out that routers not only separate the IP domains, but also the collision domains. Many people don't know/realise this fact, although they should be able to connect the dots (it's better not to assume).

Routers are OSI Layer 3. Switches and Bridges are OSI Layer 2. NICs, hubs, and repeaters are OSI Layer 1. If there is anything wrong in those statements, then Cisco is wrong, but that's not the case, so this article is not incorrect in stating the OSI layer is 3. I would suggest that the article is not misleading, but that the TCP/IP model is misleading; my proof being that most University grads in the US who take senior level networking courses do not know that the things at BestBuy are merely "intelligent" switch and NOT routers, no matter what the box says. Daniel Owens (Dohedo) 18:14, 23 February 2007 (UTC)


 * Cisco distinguishes between L2 and L3 switches in its product line. Typically, a L2-only switch has a cheaper software executable (e.g., SMI versus EMI). All switches are not L2. Cisco training, as opposed to Cisco engineering, does not try to coerce routing architecture into the basic OSI Reference Model. For anyone seriously trying to use OSI architecture, which the IETF does not, you would, at the least, need to supplement the Reference Model with the Internal Organization of the Network Layer and the OSI Routeing [sic] Framework documents.  I suggest RFC 1812, which is getting a bit old, is far more definitive about the real-world, vendor-independent definition of routers than any OSI document.Hcberkowitz 02:54, 26 June 2007 (UTC)


 * Excuse me But the OSI model is a MODEL that defines the world as the writers of that spec perceive the world into 7 layers.  It does not mean it is the only MODEL!  TCP is one of the main protocols in TCP/IP networks. Whereas the IP protocol deals only with packets, TCP enables two hosts to establish a connection and exchange streams of data. TCP guarantees delivery of data and also guarantees that packets will be delivered in the same order in which they were sent.   The Internet was built on the TCP/IP model  OSI refines it into its own spec.  If you want to write about these things in the article, Great!  But it does not belong here.  There should be an article for the OSI model already, for example.  You do not talk about models or network structure in the article that defines what a router is.  A router connects networks together.  Thats it.  You dont define protocols like RIP or ICMP here, you just use wiki links to point to these other articles.  The average user reading wiki does not have an engineering degree.  Also the images need to be realisic, they have been boldly replaced.  IMHO, tcp/ip is much more important, its how things work but not even that belongs in this article. --User: (talk • contribs • count) 21:48, 22 June 2007 (UTC)

What belongs on this page
Update the article Router if you are updating infomation about Routers, NOT routing protocols. Update the article, Routing protocol if you want to update information on the protocols routers use.

Finally update the article, Routing, if you want to update this topic without discussing the protocols behind routing.

Please remember: in the beginning there were no routing protocols, everthing has static maps. Splitting the articles in this manner was the best way to avoid confusing the average user. Remember, Wiki is an encyclopedia, as such it is not a how-to guide or a user manual. the articles are connected with Wiki links so if I usere wants the detailed information, it is only a click away. Please do not place the wrong information in the wrong article or your edits will be reverted. --User: (talk • contribs • count) 11:50, 26 June 2007 (UTC)

Pictures
This articles needs pictures for real routers, not just home ones. St.isaac 21:27, 2 September 2006 (UTC)

My home router seems very real to me. --Marco 17:13, 19 October 2006 (UTC)


 * Although "virtual router" is (forgive me) a very real concept, especially as one of the two major architectures for provider-provisioned virtual private networks. Just to confuse things, VRRP, the Virtual Router Redundancy Protocol, is a very different idea. Virtual routers, in the VPN sense, are multiple software constructs in a single physical box. VRRP and its relatives are link-local failover protocols that let hosts see a single local router, but that actually might be a set of physical routers in failover relationships. Hcberkowitz 13:12, 26 June 2007 (UTC)

I agree with St. Isaac, a Nat box is only allowing multiple computers access to the internet but does not route packets. A true router routes packets, so we should have a picture of a real router to limit confusion. Cdscottie 19:21, 29 October 2006 (UTC)

Here is a. It is used as a edge router at the ISP where I work. Is it real enough? If it is and the picture is bad, I can go take a better one. Jonathan Auer 22:25, 29 November 2006 (UTC)

Could always just dig up a Cisco 2800 series router. Cdscottie 2 December 2006

I think the current Cisco 7200 router picture is very representative of routers used in data centers. What I find nice about the current picture is that it shows the large number of interface types possible. Kgrr 20:59, 13 December 2006 (UTC)

This article does need more about non-domestic routers. Secretlondon 19:54, 22 February 2007 (UTC)


 * Added new photo of Ent class router and basic 1800 router covering a wide range of routers. Routers in the home are most always DSL Routers to avoid confusion, there is a seperate article stub for them.  Remember DSL routers are not DSL Modems and are not DSL bridges DSL modems are DSL Bridges. so you should not include home or res. routers here I am just adding a wiki link --User: (talk • contribs • count) 20:24, 23 June 2007 (UTC)

Red link in edit #141103024
Regarding the red link in my comment for my recent edit, #41103024: I must've inadvertently copied the page title incorrectly. The reference should link to Guide to layout. Sorry about that! – Mipadi 04:17, 28 June 2007 (UTC)

Created FAQ
Created FAQ section to deal with future issues I see comming as we move to the next set of articles and issues. --User: (talk • contribs • count) 14:58, 29 June 2007 (UTC)

In the interest of not being overwhelming...
I strongly suggest not putting "routing table" here. Instead, speak in general terms that a router has to have a "map" of the network, perhaps with "directions", as a simple analogy. Then, establish that the control plane builds the map and writes the directions, and the forwarding plane is the "traffic cop".

Routing table would then be subordinated to control plane and FIB to forwarding plane. I recognize the need to have nontechnical introductions, but, to go much beyond the concept of "map" and "directions", the subject of routing tables and the associated control gets quite technical, and there is a huge amount of misinformation circulating about it. Howard C. Berkowitz 21:08, 2 July 2007 (UTC)


 * There seems to be various discussion of merges between routing-related articles. Maybe we can consolidate the discussion here?


 * Regarding Forwarding Information Base: I see that Forwarding Plane and Control Plane are not in fact stubby "little articles" as I might have implied. So I agree with Hcberkowitz that a fine first step is to merge Forwarding Information Base into Forwarding Plane, and Routing table into Control Plane.


 * As a second step, I think it may be preferable to merge Forwarding Plane and Control Plane into Router since they are core topics for a router. I think there is value to having all the core concepts in one article, which provides more structure and context for the reader so it's clear how all the parts fit together. However, one valid reason not to do this could be length of the articles.  At their present length, I think they're worth merging. The rough guideline for maximum article length is 30-50 KB; router is currently at 14 KB. I see though that Howard has been working extensively on all these articles so perhaps he can comment on their possible future sizes. (I don't really understand the point about having too much technical detail, unless you are also in effect referring to article length.) --Nethgirb 12:34, 3 July 2007 (UTC)


 * Looking at all of this and looking at it with an admin; is is possible to leave it all alone as is. After long discussion, I would leave routing table alone as well. It is not necessary to merge the documents so long as they have an easy to understand intro that anyone can understand.--User: (talk • contribs • count) 22:32, 6 July 2007 (UTC)

The "Core" Problem
I'll try one more time to express a general consensus definition of "core" in a provider sense, which I've explicitly said is different from an enterprise sense. In the provider context, it is somewhat reasonable to talk about a "core" of autonomous systems: Again, if the idea of the "Internet core" is meaningful, it should be possible to find authoritative citations that use that terminology. It should also be possible to characterize, with citations, the definition of a "core router", if again that is a meaningful term. Howard C. Berkowitz 14:05, 20 July 2007 (UTC)
 * 1) As distinct from core routers. A provider core router is not exposed to the outside world.
 * 2) The definition, even at the level of granularity of AS, is much more statistical than topological. We do speak of the "skitter core" based on CAIDA's measurement of the highest-volume AS or set of AS.
 * 3) In the past, a "core" provider was default-free, but router hardware has advanced to a point where even enterprise border routers can hold multiple views of the global routing table.
 * 4) Internal to a provider, routers that are topologically part of the internal core may not even run BGP or have an Internet routing table, as in a RFC 2547 "P" router.


 * Howard, I think you need to have a look at the article I pointed you to. You still made no comment to me on it.  Core router this is what a core router was as defined by the editor of that article.  (I did not write it)  I am not saying this is true now, I am saying what was needs to be preserved.  Are you trying to tell me that this editor is wrong and this article is wrong?  I really dont like being in the middle of editors but if you tell me the def is wrong I will research to verify... --User: (talk • contribs • count) 02:01, 22 July 2007 (UTC)


 * The Core router points back to Internet backbone. The latter article says: "The term "Internet backbone" is now sometimes loosely used to refer to the inter-provider links and peering points. However, with the universal use of the BGP routing protocol, the Internet functions with no single central network at all." That is my chief point, that there is real value in pointing out there is no true core.  The core router article doesn't clarify that the referenced internet backbone article essentially says there is no backbone any longer. As long as articles do not make it clear that the Internet hasn't had an identifiable physical backbone for quite a few years, I'd call them significantly misleading. As long as Core router fails to make this distinction and perpetuate a myth, I'd call it wrong. Yes, telephone switching engineers still use the insider shorthand of class 4 and class 5 switches, but I wouldn't say that in an introductory article. Mentioning classes 4 and 5 bring up the obvious question of classes 1 through 3, which were part of the AT&T monopoly switching structure which hasn't existed since the late seventies.


 * The backbone article still perpetuates the misconception, especially for North America, that the majority of traffic is interchanged at multilateral peering point. It's not. It's interchanged through bilateral links between members of what is called the skitter core, which is not a topological but a statistical definition. I know of no diagram on which one can point to the skitter core, because it is made up of numerous interconnections of major carriers, often comprising multiple autonomous systems, literally across the world. Skitter core visualizations are traffic hierarchies that have very little to do with physical topology, and thus it's misleading to speak of core routers as suggesting they are in the role of provider border routers. The highest speed routers are likelier to be "core" in the sense I described of enterprise core, or, in a VPN sense, VPN P routers that may very well not run BGP. Some very large provider architectures, incidentally, treat the public Internet as one more VPN that goes onto MPLS through a P router.


 * Best current practice is to do 100% deep packet inspection of interprovider traffic, unless you are enormously sure that the other provider has already done that. The 40 Gbps links (e.g., Cisco CRS-1, Juniper T-640) are more likely to be internal to large providers, or to be used very, very carefully on trusted interprovider links.


 * For the typical high-performance interprovider link, Cisco's recommendation is to use what they call router clusters. A router cluster typically involves a pair of distributed forwarding routers (e.g., GSR 12000) linked by a high-performance layer 3 "switch" such as a 6500 or 7600 (essentially the same device, but the "switch" version has the boards in a typical enterprise air conditioning configuration but the "router" version has them at carrier right angles. The "router" has a few more interface types).


 * Router clustering recognizes that to do deep packet inspection, you run out of inspection cycles before you run out of forwarding cycles. You start the first security inspection on the ingress forwarder of the outside router, and throw away some packets. Next, you do more inspection on the independent processing of the egress forwarder of that same router. The intermediate device lends itself to attachment of firewall and performance analyzers, as well as doing rate limiting. Next, more inspection and discarding happens on the ingress forwarder of the inner router, and the last stage (perhaps NetFlow recording) on its egress forwarder.


 * Several router clusters might well feed an internal maximum forwarding performance router, but note that the internal router, arguably a "core" internal to the provider AS, not the Internet, is not part of a homogeneous "internet core". When you speak of "preserving", in my opinion, there must be a solid and accurate starting point, else errors propagate everywhere. What is to be the starting point? Router, which then might point to the subclass of Core router? Must Internet backbone be preserved when Internet architecture would be far more accurate (and has been so for decades)? Howard C. Berkowitz 03:05, 22 July 2007 (UTC)

Historical note
There is some uncertainty about the exact date of the first multi-protocol router, since much early documentation has been lost.

In January, 2001 I was contacted by a reporter, Peter Carey of the 'San Jose Mercury News', who wanted to do a story on this, and we spent a lot of time trying to figure out who did what, and when. As part of that investigation, I spoke at length on the phone directly with Bill Yeager. Neither one of us knows for certain when our multi-protocol routers first transmitted packets. (I do have a piece of paper showing that my single-protocol router first sent packets on about 13 March, 1980, but that doesn't help.) Neither knew of the other's work at the time, we independently invented the idea.

As best we can tell, mine was running around the end of 1980 / start of 1981, and his in late 1981. His interview with Network World actually does say late 1981 for his, if you read it carefully:


 * Initially, the code routed Parc Universal Packet (PUP) ... Late in '81 my boss said, 'IP is coming down the pipe. Figure out what you can do with it.' So I put a little IP router in ... But we were ready by '82

(I have a complete file of email between us, and numerous other people, including Bill Yeager, available to any serious historian who wants to look at it. I don't want to put it openly online as it contains a lot of private email.) Noel (talk) 18:27, 11 August 2007 (UTC)

First router
I removed the text:


 * The first modern (dedicated, standalone) routers were the Fuzzball routers.

from the article because it's wrong. The first routers (in the sense of packet switches that work off internetwork-level headers) were either:


 * The first BBN routers (which were done in BCPL, iirc, although it may have been Bliss-11) under an OS called ELF, running on PDP-11/40's, or
 * The earliest PUP routers at Xerox PARC.

According to The Fuzzball, by David L. Mills, the first Fuzzball router was around 1977, and those other two were almost certainly earlier - circa 75 or so. Noel (talk) 03:43, 2 September 2005 (UTC)


 * The article text is clearly wrong, but the alternatives suggested are not right either. Assuming the ARPAnet is the first packet network -- which sounds right -- the first routers obviously would be whatever ARPAnet first deployed -- which is the IMP, in 1969.

I confirm this IMP was the first router. --User: (talk • contribs • count) 14:36, 22 June 2007 (UTC)


 * Nope. The IMP was a network-layer device, whereas routers are internetwork-layer devices. I.e. routers were conceived to connect networks with dissimilar network layers, e.g. point-point links and LAN's. You will find this all laid out quite clearly in the original paper which posited routers (or 'gateways', as they were called back then): Vinton Cerf, Robert Kahn, "A Protocol for Packet Network Intercommunication". Noel (talk) 18:47, 11 August 2007 (UTC)


 * I've tried introducing X.25 switches, and I like your distinction between network-layer and internetwork-layer devices. Clearly, a transient virtual circuit identifier is not globally significant. Incidentally, I'm trying to remember the designers; Noel, your memory may be better than mine. Holger (Nielsen???) was the hardware designer for the multi-6502 Telenet 2nd generation. Again IIRC, John Holmblad led the software team. I knew them, but only went to work for Telenet a number of years later. Since I'm in mid-move with files in storage, I can't lay hands on some of the early Telenet documents I may still have. Howard C. Berkowitz 21:40, 11 August 2007 (UTC)


 * I don't know the details of Telenet, alas; my X.25 expertise is limited. Also, your addition has a minor error: IMPs were not pure connectionless devices, since in the ARPANet reliability was assured by the network, not by the hosts. They didn't have full connections, of course, but something half way in between. (CYCLADES was different, of course, which is where TCP/IP got the idea - too bad CYCLADES isn't as well known as it deserves to be; outside the ARPANet, it's the most important technical predecessor to the Internet.) But your general point - that routers don't do connections - is a very important one, and I'll move it to the place where it does apply. Noel (talk) 03:19, 12 August 2007 (UTC)


 * Thanks. Lots of things, as you well know, don't fit neatly into simple definitions. Shall we revive the locator versus identifier argument vis-a-vis routers? :-) RSVP state in IP routers still is connection-oriented, as is (G)MPLS. MPLS labels also have a strong relationship to transient, link-local identifiers.
 * For information, Telenet, internally, used a slightly tweaked X.25 protocol, which included X.121 source and destination addresses, and the addresses of one or more packet switches that might be boundary routers between different "routing" domains. Tymnet, on the other hand, was quite different. Tymnet was built around the concept of a central route server, which was reliable in the sense that it was implemented as a set of redundant servers with VRRP-like failover. Perhaps I'm reaching, but if I were to draw an analogy to Tymnet call setup, it reminds me of NHRP.
 * I've never been completely comfortable with calling a modern router stateless, unless, and perhaps not even there, we are only discussing the forwarding plane. Certainly, routing protocol peerings/neighbor relationships/adjacencies are stateful, and the RIB and FIB do have a certain level of statefulness.
 * Apropos of your point about moving to "network switch", I'm uneasy, as "switch" too often is marketingspeak, or shorthand for "things that aren't quite routers or bridges". I'm beginning to wonder if there needs to be an article, perhaps disambiguation, for a range of connection-oriented relays (deliberately using the generic OSI Routeing Framework term), from PSTN CO switches to MPLS (at least) LERs to MPLS-enabled IP routers to ATM switches. Note that while some of these instances, in spite of the all-too-common tendency to try to coerce functions into the original OSI model, not counting the Internal Organization of the Routing Layer, may be called "L2.5", "L2", or "L1". What I've tried to exclude is connectionless (in the forwarding plane) bridges. Howard C. Berkowitz 14:52, 12 August 2007 (UTC)


 * Yes, RSVP does have state, but it's not related to the transport layer - if the RSVP state is lost, the TCP connection doesn't suffer (unless packets cease flowing entirely, of course). It's a bit tricky to find exactly the right word, or short phrase, to accurately and exactly describe how it is that routers aren't involved in the end-end communications. Perhaps "connectionless" is the best single word after all. (I guess IMPs were slightly connection oriented, in that the Host-IMP protocol used links, in particular, messages sent on a single link would be delivered in order, IIRC, and NCP connections mapped directly from sockets to links).


 * Not just RSVP, but dropping routes on not hearing hellos, MPLS recovery, etc., don't have simple and clear terminology with respect to the broad concepts of state and connection. The effects, for example, of losing an RSVP reservation in VoIP may manifest itself at the level of SIP, not pretty from a layering standpoint. One approach, which may be getting too obscure for Wikipedia, is separate consideration of state in (at least) the forwarding and control planes. FORCES just scratches this, not considering service planes and only abstractly dealing with stateful failover in multiprocessor routers or groups of routers. While it isn't that common, just as a thought, where does ICMP Redirect fit? Howard C. Berkowitz 14:10, 13 August 2007 (UTC)


 * Speaking of which, you have a good point that there is a range of "connection-oriented" packet-switches, which I would arrange from things like X.25 at one end, to ATM at the other, using the metric (which seems the most natural one to differentiate along, to me) of how much of the tranport-layer functionality is in the packet switch. E.g. X.25 and IMPs do end-end reliability in the packet switches, and ATM doesn't, etc.
 * Your Telenet and Tymnet information is most interesting. Can I encourage you to add it to those articles directly? As you can see, we have articles on both of them.
 * First pass done for Telnet last night. Unfortunately, I'm in the middle of a move, with hard copy documents in boxes in storage. I'm afraid I may have thrown out some of the detailed implementation manuals. There's a Wikipedia problem here of referencing, when, variously, I did some of the work myself, or can only say, from having worked there, "Yeah, Holger Opderbeck invented that, but I don't know if it was just something we all knew or it was written down in any one place." I may have found some patents on bus arbitration in his multiprocessor switching engine, but they don't deal with the overall architecture.


 * I've realized that a term first, AFAIK, in the OSI Routeing Framework, "quasi-static routing", is useful but very rarely used these days. It's a superset of the more common term "floating static route." Howard C. Berkowitz 14:10, 13 August 2007 (UTC)
 * Also, I didn't move the X.25 content to network switch, but rather to packet switch - which I think you will agree is a fairly well-defined term! It seemed a better fit there, in a "History" section, than it did in Routers. Noel (talk) 12:46, 13 August 2007 (UTC)

This information was removed from the artilce because it was not correct. Even though this happened it happened much later and yes it was the first multiprotocol router, Leave here as a reference as I read notes that say the History section was too complex. Could someone advise? --User: (talk • contribs • count) 14:58, 22 June 2007 (UTC):

This data has now been re-inserted into the article in the proper place. This article still needs help. Some of these external links should be references. Article cleanup needs to continue. Out of time today. --akc9000 15:18, 22 June 2007 (UTC)


 * I don't think talk pages should be the place where "extra" material is kept. But I agree that the text you clipped was a bit much, especially the asides about C compilers.   It certainly is nowhere near being too long.   Paul Koning 15:11, 22 June 2007 (UTC)


 * Hi Paul Thanks for your help, Yes. I only dumped the data here for about 10 minutes while I was working on the sites, incase someone else was looking at the article and an edit confict happened.  Your trim is what I was going to do so thanks!  This article needs alot of help so I am glad your here working on it with me. --User: (talk • contribs • count) 21:32, 22 June 2007 (UTC)

I added some text to the article about the origin of the concept of a router (in the Cerf/Kahn paper), and also about the first BBN router. I forgot to add something about the PARC one; I'll try and dig up a reference and do that now; I'll also try and more accurately reflect the roots of the gateway idea, which weren't specific to the DARPA project, but went back to the earlier INWG work. Also, see the section below, Talk:Router, for a note about multi-protocol routers. Noel (talk) 19:01, 11 August 2007 (UTC)

"You see an exit..."
I've gone ahead and removed the following, initially as it appeared to be original synthesis. While it's now claimed that the text is from elsewhere, this presents an entirely different problem&mdash;uncited material taken from other sources is plagiarism and a potential violation of copyright. So one way or the other, it's got to stay out without a source cite. If put in, it would need significant cleanup, since we never use second-person ("you") in articles, unless it is a direct quote and can be quoted and attributed. (Even then a paraphrase might be best though.) Seraphimblade Talk to me 03:08, 29 August 2007 (UTC)

Reference Section in Disarray
Right now the reference section is a complete mess. I'm not sure why it's like that, but I feel as though it should be cleaned up. I would go ahead and do it, but I'm not sure if there's a reason for that, or if it's simply a mess. —Preceding unsigned comment added by Kingdom Alaska (talk • contribs) 01:31, 13 December 2007 (UTC)


 * OK, fixed. There was a missing close ref at the end of the first reference. Howard C. Berkowitz (talk) 02:30, 13 December 2007 (UTC)

Other brands
Does anyone have pictures of other company's products that would fit in this article? 4 pictures is good, but all 4 from the same company is not really desirable. Paul Koning 10:52, 17 July 2007 (UTC)

Hmm, but Cisco is the industry leader, and far larger than any other manufacturer. —Preceding unsigned comment added by Thesydneyknowitall (talk • contribs) 14:35, 29 December 2007 (UTC)

I Agree. found this on wiki  69.140.191.131 18:42, 18 July 2007 (UTC)

Great, but now if you use this photo you must say something about it, or it will get deleted for being decoration. If you wonder why I have not been editing it is becuase there has been a debate over the Cisco images. Because they were not 'free'. The other reason why there were only Cisco images is because the VP granted GDFL for these images.


 * Please do not put multiple pictures of the same thing in this article. The second image has been removed.  The information also has been removed from the simple intro.  Remember the simple intro is a simple intro.

Must be mentioned in the article
I think it is good to have another brand but it just cannot be there. There has to be something mentioned about it in the article. Like what it is? What does it do? In other words say something / anything about it in the article. --User: (talk • contribs • count) 02:48, 19 July 2007 (UTC)

People, if you want to add picture of routers great but this is a photo of an Metro Ethernet Routing Switch 8600. It is a switch, that can route but it is sold as a Metro Switch not a router. So why not put the picture in the Network switch article? --User: (talk • contribs • count) 01:46, 20 July 2007 (UTC)


 * You are perpetuating the myth that there is meaning, beyond marketing speak, to "switch". Given that I was product line manager for routing protocols, in group that was called "carrier routers", I think I have just a bit of familiarity with the nature of the 8600. We were responsible for the 8600 as well as the(discontinued) Versalar 15000 and (cancelled) Versalar 25000 routers. All ran exactly the same BGP as the 8600. The only difference in the routing is that the 8600 of the time had a smaller routing table, and only Ethernet interfaces rather than SONET/SDH, etc. There isn't a hope of ever getting a reasonable definition of routing and routers, and, for that matter, bridging and bridges, if "switch" is treated as anything more than a redirection entry to a reasonable term. "Switch" comes back to a marketing slogan (Synoptics/Bay, depending on where you were in the merger history), aimed against Cisco, "Switch if you can, route if you must." At roughly the same time, Cabletron called all its devices "hubs", even though the "hub" was a hardware shelf that could have layer 1 repeaters, layer 2 bridges, and layer 3 routers in it. Indeed, the "router" card in the Cabletron "hub" was long a Cisco IGS card licensed to Cabletron. Howard C. Berkowitz 02:29, 20 July 2007 (UTC)


 * For quite some time, the top of Cisco's line was called the GSR, numbered the 12000 series. GSR stands for "gigabit switch router". Would a picture of a 12000 be a picture of a switch, a router, neither, or both? Howard C. Berkowitz 02:29, 20 July 2007 (UTC)


 * I would say both :-) As I said I have no problem with the photo if there is a consensus on it.  --User: (talk • contribs • count) 13:11, 20 July 2007 (UTC)


 * To add to my clarification of a caption of Cisco's current top-of-line, the CRS stands for Carrier Routing System. Is "routing system" significantly different than "router"? As in all network definition, "it depends". There is an increasing need for "provider border" routers, as opposed to "provider core" routers, to do very detailed examination of packets to protect against increasingly subtle security threats. One approach that has been used is to set up, either in one chassis or a set of chassis, a pipeline of forwarding processors that also do security inspection. Howard C. Berkowitz 14:05, 20 July 2007 (UTC)

Street analogy
While there is a comment (within the source code) that the street/freeway analogy is confusing, there seems to be no attempt to improve it. Can someone think of a better analogy? My solution would be remove the analogy entirely. It suffices to say that switches know only of limited (their own) IP addresses while routes maintain a full list (routing table).
 * You know, I just clicked on the "Discuss" tab because I wanted to congratulate the author of this paragraph, then I've seen your message. I find the analogies between the street and the switch, and between the IP and the home adress very clear. It gets a little more complicated when it introduces on-ramps and freeways. There are probably better words for this. But please keep the first part of this paragraph as it is, I find it written in a very understandable way. My understanding of what is a router and how different it is from a switch has improved thanks to these few lines. --CutterX 21:36, 7 July 2006 (UTC)


 * I found the analogy to be confusing, and not particularly helpful. Rosensteel 13:05, 2 August 2006 (UTC)


 * I agree, I had to read it several times and use previous knowledge to make sense of the analogy. I'd suggest clean it up or remove it. Jaws87 14:28, 8 November 2006 (UTC)


 * In my PC Config & Repair class, my professor had a good one: let's say that he slept with the girlfriends of all the guys in some bar. As they close in to kick his ass, he points away and shouts, "What's that?!"  As they all turn to look, he jumps over the bar and turns out the lights.  Now if he asks, "Hey, where's Bob," Bob replies, "I'm over here!"  Some other people around Bob would say "Hey, I heard Bob."  The router is the guy by the door.  We can go to the guy at the door and say, "is Jim here?"  The door guy looks around and says, "no, he's not in here.  I can see him outside, though."  I don't know if that was exactly it, but that's the gist of it. --CCFreak2K 10:36, 5 December 2006 (UTC)


 * I merely fixed the analogy that had been in use before (which was totally messed up). I have used the streets and freeways analogy in many places because the media labels the Internet as information highways or the "infobahn".  Two points that are important to routers: they join networks and they have a routing table.  Whatever analogy you end up using, it needs to make these two points clear.  The street analogy makes sense since it does join two road network and the routing tables are the street signs that have the block numbers written below the street name.  I'm open to replace the analogy if you can find a better one.


 * By the way, the bar example does not make a huge distinction between all the guys in the bar versus the guy standing at the corner (who is the router) pointing to where people ran off to.


 * I always used the analogy of a traffic cop at an intersection, directing the cars (Packets).


 * Kgrr 21:10, 13 December 2006 (UTC)


 * The analogy doesn't make any sense at all. Switches -- as in "Layer 2 switches" -- and routers both switch packets based on their address.  They differ in a few ways: (1) for routers, the addresses have structure, and that structure is used in various aspects of what routers do.  (2) routers communicate detailed topology information amongst themselves, while switches communicate very little (basically, just the spanning tree).  (3) routers route packets according to some "best path" rule, while L2 switches use the spanning tree.  (4) routers modify the packet (specifically, the hop count or equivalent in the L3 header) while switches pass the packet through without change.


 * None of these differences relate in any way to anything suggested by the streets analogy.
 * Paul Koning 00:46, 2 June 2007 (UTC)


 * Hopefully my corrections make alot more sense. --User: (talk • contribs • count) 11:54, 25 June 2007 (UTC)


 * Houses are the PCs; their Street Address are the PC's IP address. Houses in a particular area share a street. PC's share a "switch" (At this point, ignore L1,L2,L3). All cars leaving their houses drive onto their shared street like all PC bandwith traffic leaving their PCs go onto their shared switch. However, 1 car may simply drive down the same street to a convenience store to buy something, while another car needs to work. The 1 car going to the conv. store is like a PC that goes to the switch to connect to a server. The other car is like the PC that has to goto the router. At the intersection (the router), the PC simply goes onto another network (another street address). What's confusing in the analogy is that the other network is a private network (10.x.x.x) This is okay, but to the car staying in the same general area with privates streets. They can only travel within their own neighborhood; their own network organization. What about the car that wants to go from Chicago to New York? The analogy should include some reference to a router that allows travel out of the private network. —Preceding unsigned comment added by Binaryoptions (talk • contribs) 11:16, 25 April 2008 (UTC)

as
as —Preceding unsigned comment added by 221.134.144.14 (talk) 20:35, 8 October 2008 (UTC)

about router
whose first developed router and which advance router currently use? —Preceding unsigned comment added by 59.96.159.128 (talk) 11:51, 18 November 2008 (UTC)

rewrite of intro needed
The intro needs more of a lead into the general topic of networking and discuss the differences between hubs, switches, etc... Saying that a router generally has specialized OS, should be discussed later. What do you think? Thx, Daniel.Cardenas (talk) 17:31, 20 November 2008 (UTC)

Bridge
Should "bridge" be mentioned among the opening paragraphs, along with "router", "switch", "interface", and "network hub"? The fact that something (in this case, a "switch") doesn't have a precise definition shouldn't prevent us from attempting an approximate one. Perhaps it would be best to avoid the word altogether, except to mention that it's a misnomer, and what it usually should be. Unfree (talk) 21:06, 13 December 2008 (UTC)

Similarity to computer
I see here, "Routers generally contain a specialized operating system ..., RAM, ..., and one or more processors, as well as two or more network interfaces. Except for multiple network interfaces this is typical of an embedded computer." I wonder, "Why 'embedded', and why 'except for'? If the intention is two-fold, to show what routers contain, and to liken them to computers, then perhaps "network interfaces" ought to be left out of the initial sentence, and mentioned after the one likening routers to computers.  That way, the "except for" could be avoided.  I don't see why "embedded" is even mentioned. Unfree (talk) 21:22, 13 December 2008 (UTC)

Cleanup
Update the article Router if you are updating infomation about Routers, NOT routing protocols. Update the article, Routing protocol if you want to update information on the protocols routers use.

Finally update the article, Routing, if you want to update this topic without discussing the protocols behind routing.


 * If I may, let me suggest some alternate terminology, which is used in the IETF and other specific forums, such as the Routing Research Group of the Internet Research Task Force (IRTF). Unfortunately, there's a great deal of confusing information that comes from essentially marketing terms such as "switch". There is other confusing information that gets in from training and certification material that was written by pure education people extracting from engineering documents, rather than engineers.


 * I'm sure some of this should be in bullet form, but I'm not yet comfortable with that aspect of Wiki formatting. Forgive me, and feel free to clean up.


 * I'd like to suggest that the three topics here really are routers (the devices that do, in general, both packet forwarding and control of the forwarding intelligence), the forwarding plane of routers (the act of moving packets from interface to interface and from router to router), and the control plane of routers (the intelligence behind the specific forwarding choices, which includes, but is not limited to, routing protocols). Yes, there are other nuances here, but these are the three main building blocks. For example, it is an implementation decision in designing a router whether to use the same processor for forwarding and control (usually the case in low-end routers), while in a high-end router, there is usually one control processor (with backups), and many forwarding elements that can contain multiple processors and ASICs.


 * While it is getting a little dated, RFC 1812 is the current "router requirements document", which defines, from the perspective of the IETF generally, what a router does. Paraphrasing, a router is a device that uses principally IP information to make decisions regarding the interface(s) out which to send a packet received on another interface. I am very deliberately saying "IP" rather than "network layer" or "layer 3", because, in the real world, there's no new development on non-IP routing. IPv4 and IPv6, yes. MPLS and GMPLS are really beyond the scope of this discussion, but don't contradict it.


 * A router (i.e., a physical piece of equipment) has two basic functions, often called planes: the control plane and the forwarding plane. For further vendor-neutral discussion, see the IETF FORCES working group, http://www.ietf.org/html.charters/forces-charter.html, which deals with the actual separation of control and forwarding elements. Higher-end routers, which some vendors call layer 3 switches, for example, may have one active control element but many distributed forwarders; the distributed forwarding engines get network topology information from the control processor. High-availability routers may have multiple control elements, but, in present designs, only one is active and the others are for various kinds of failover.


 * The forwarding plane has the highest processing speed requirement. In high-performance forwarders, there are one or more data structures, per physical forwarding element, that are abstractly called Forwarding Information Bases (FIB). A FIB is optimized for making extremely fast lookups of destination addresses and other information used in forwarding decisions. Low-end routers may not have a separate FIB.


 * The control plane needs intelligence and large memory more than it needs raw speed. Routing protocols run among control elements, and, along with static configuration and hardware status, build the "routing table", or, more formally, the "Routing Information Base (RIB)". Implementation-dependent mechanisms cause FIBs, when separate, to be populated from information in the RIB. The RIB is optimized for updating by routing protocols rather than fast lookup. RIB optimization is especially significant for Internet core routers that may have high hundreds of thousands into low millions of routes, some of which are backups and do not appear in a RIB.


 * For completeness, there may be a services plane that handles additional functions such as encryption, deep packet inspection for security and QoS, forwarding decisions based on higher layer information, etc. Howard C. Berkowitz 13:03, 26 June 2007 (UTC)


 * Re "what is a router" and RFC 1812: Please do not limit the discussion to IP only. It may well be that "there's no new development on non-IP routing".  But that doesn't mean non-IP routing isn't notable and worth discussing.  TCP/IP is not the only protocol stack in the world, nor the best-designed one.  There are things to be learned from discussion of other protocol stacks.  And there are historical connections too; for example, OSPF was derived from ISO IS-IS. Paul Koning 13:55, 26 June 2007 (UTC)

Please remember: in the beginning there were no routing protocols, everthing has static maps. Splitting the articles in this manner was the best way to avoid confusing the average user. Remember, Wiki is an encyclopedia, as such it is not a how-to guide or a user manual. the articles are connected with Wiki links so if I usere wants the detailed information, it is only a click away. Please do not place the wrong information in the wrong article or your edits will be reverted. --User: (talk • contribs • count) 11:56, 26 June 2007 (UTC) Placed on top out of order so it may be noticed --<span id="" class="plainlinks" style="color:#002bb8">User: (talk • contribs • count) 11:56, 26 June 2007 (UTC)

I'm flagging this article as in need of cleanup for several reasons, primarily because I find it hard to understand, even though I already know what a router is. (Note that I'm splitting this message into three or four parts to facilitate discussion of the different points made)--Verbatim9 01:45, 4 February 2007 (UTC)


 * (please discuss whether this article should (or should still) be flagged here...once there's discussion this text can be deleted)


 * It seems to me that the article says next to nothing about what a router is or does. Most of it is simply a list of vendors, which should be deleted. Paul Koning 18:41, 15 April 2007 (UTC)

Function section cleanup
In the "function" section, the article gets into what a router isn't too soon, before it has really finished describing what a router does. Contrasting a router and a switch may be useful in explaining what makes a router a router, but after the reader has been given a grasp of what a router does. Also, I think some comparison is in order...switches are after all hubs that perform a kind of low-level routing, determining (whenever possible) what computer each packet is for, and sending that packet only to the appropriate computer...it would be good to explain that (in a few words) before contrasting it with the higher-level functionality of a true router, and offering analogies. --Verbatim9 01:45, 4 February 2007 (UTC)


 * (please discuss/propose changes to the Function section here)

I agree that a comparision is in order; switches should be compared to hubs in the switch section, routers to switches in the router section. What I'm not a fan of is your description of a switch. Switches don't perform routing even "low level" routing; switches merely foreward information to a port that they know the destination device is on; the key being that they foreward, not route. Other than that, I agree with you. Daniel Owens (Dohedo) 18:55, 23 February 2007 (UTC)


 * Yeah, I didn't think I knew enough about switches (among other aspects of the article) to describe the relationship correctly, which was part of why I posted on the discussion page, suggesting a direction for improvement, rather than editing the article myself. Having read up on them a little more, I see your point...they both keep routing/forwarding tables used to send packets/frames in the right direction, but a switch/bridge leaves the frame unchanged, whereas a router replaces the frame, and re-packages the data as well when necessary (e.g. due to NAT or packet size/format changes between the networks). None of that is reflected in the current analogies... Verbatim9 20:31, 7 March 2007 (UTC)

"Switches don't perform routing" - What about layer 3 switches? do these perform routing? Or is it still just forwarding to an address? RoninNZ 23:59, 15 March 2007 (UTC)


 * The problem here is "switch" is a marketing term rather than a technical one, and is not used in any serious work in routing architecture. With that caveat, products called "switches" rather than "routers" tend to be optimized for principally Ethernet interfaces. Within Layer 2, switches started differentiating from classic bridges by using microsegmentation, assigning one device per physical port rather than having classic shared-medium Ethernet. As long as the ports are full-duplex, this technology avoids collisions. "layer 3 switches" still tend to be optimized for Ethernet interfaces, and, in the more edge devices, have small but fast forwarding information bases, often with limited routing protocol support.Howard C. Berkowitz 13:08, 26 June 2007 (UTC)

I cleaned up some obvious grammar in the Function section, as well as qualifying a reference to IOS (which is the Cisco-specific router OS). I don't really like the way the routers/switches comparison in written; it reads too much like opinion, and there's enough difference between routers and switches that I think the current wording confuses the issue. Sonicforest 15:27, 11 July 2007 (UTC)


 * Useful changes indeed. Apparently, we were editing simultaneously, and it's the first time I got an edit collision. My changes may not have merged gracefully.


 * Deep in my heart, I wish we had no "switch" article, and just stayed with repeater, bridge, router, and so forth. Part of the problem is a past marketing slogan (Synoptics? Bay?) "switch if you can, route when you must", but the implied speed difference really doesn't apply any longer. I can speak plausibly of edge switches and campus switches as LAN-optimized subsets of bridge and router. There are exceptions, but the majority of "switches" actually route. The bridging solutions, in a campus of any real size, get very ugly although they are sometimes needed (if, for example, you don't implement mobile IP and you want a flat wireless network that works anywhere in the campus).


 * It is probably worth putting in that most "switches" have operating systems. You were correct to change the Cisco-specific reference; I'm not sure if it is a good addition or not to add another, such as Extreme's XOS. It may be too much information to speak of some of these as written ground-up and some being UNIX/LINUX derivatives. Howard C. Berkowitz 15:55, 11 July 2007 (UTC)


 * Agreed with your point about "switch". I made that same suggestion in Talk:Network switch.  Maybe it's time to make it a proposed merge? Paul Koning 20:01, 11 July 2007 (UTC)

While I do find this discussion of switch vs. bridge interesting, I wonder if the line "A function that forwards based on data link layer, or OSI layer 2, information, is properly called a bridge. Marketing literature may call it a layer 2 switch, but a switch has no precise definition." belongs in this article. This line would seem to be a better fit for Ethernet. Perhaps a brief reference and a link would be more appropriate. --Krennt (talk) 02:37, 17 March 2008 (UTC)

Types of routers section cleanup
I think there are a couple problems here:
 * The initial portion seems to go into too much historical and technical detail...how many typical encyclopedia readers know what a minicomputer was, how many have heard of ARPAnet before, how many know what an IMP is/was? How many typical encyclopedia readers know enough about PSTN hardware to recognize or benefit from a comment on their similarity to modern router hardware? I think some of this information needs to either be removed for simplicity's sake, or separated into a "history" section. --Verbatim9 01:45, 4 February 2007 (UTC)


 * With the addition of the history section, and an accompanying reduction in the density of unfamiliar terms, I think this particular issue is now pretty well resolved. Verbatim9 20:45, 7 March 2007 (UTC)


 * Routers in different locations in the network are notably different from one another...the functionality of each major type of router should be made clear (at a level of detail appropriate for an encyclopedia). Describing how each fulfills routing functions might serve to clarify what a router is. Currently the article talks about several different kinds of routers (core, edge, home DSL/Cable routers, "one-armed" VPN routers). It might be worthwhile to create subsections for each of these (though someone with more knowledge of the subject should decide whether core/edge routers are worth describing separately) under the "types of routers" section. Also, as requested below, it might be worth adding an explanation of wireless routers (both the standalone wireless routers used e.g. in campus networks, and the inclusion of a wireless network segment in many DSL/cable routers). --Verbatim9 01:45, 4 February 2007 (UTC)

While I agree that the differences in the different router types should be mentioned, perhaps a small section, I feel that most of the differences are in the setup, not the hardware (although that can depend on the manufacturer and your budget). More often than not, you know if something's a border device or not based on the routing protocols used. I feel that the depth is really more discussing what routing protocol is normally used where, when, and why. The actual physical differences in a core/edge router are normally tiny, if at all; I can stick a 7600 or 7200 anywhere on the network, but I'm more likely to put it on the edge just because it is expensive. The tasks performed are the same... it is really a matter of cost (a router, is a router, is a router in its functionality and it's hardware... more ports just means more costly). As for history, this is an encyclopedia; its duty is to present the history of things where relevent. I feel that the history shaped the devices and gave rise for the need to have routers, so I feel the history is fine. If someone doesn't want/need to know the history, they can skip the section; if it is in-depth, then the person will have the potential to learn more. That is just my opinion. Daniel Owens (Dohedo) 18:47, 23 February 2007 (UTC)


 * To take your Cisco example, there is a fairly significant difference between a 7200 and a 7600/6500, and even more a GSR 12000. The 7200 forwards through the main processor, while the other, higher speed devices generally distribute the forwarding decisions onto the line cards. There can also be differences between edge and core devices in the size of the routing tables they use. Many perfectly good edge devices (e.g., Cisco 3550 or Nortel 8600) cannot hold a full Internet routing table, much less with multiple BGP views. Hcberkowitz 02:50, 26 June 2007 (UTC)

In this section, I was confused that the Small Office Home Office (SOHO) section links to port address translation rather than network address translation. The strictly layer 3 NAT seems to fit better with the description of a router as a layer 3 device that may contain extra features. Linking to NAT would also be more consistent with the residential gateway page, which this section cites as the Main Article. However, these devices usually do include the layer 3/4 version, so the link to PAT might be more helpful to the average user. Any thoughts? Acobb (talk) 01:16, 22 January 2009 (UTC)

Non-English on the English wiki entry
Non-English on the English wiki entry - at the least shouldn't be in the top of the article. Sorry if I'm stepping on someone's toes. Snipped entry follows: 76.192.170.116 (talk) 19:26, 11 September 2009 (UTC)

FAQ
I am a layperson regarding networks and I visited the Router page hoping to glean a little knowledge. However, the purely technical nature of this entire article renders this page virtually useless to the average user. Could someone please edit this article to be more consistent with other Wikipedia content; i.e. can some please edit this into an encylopedic format rather than a technical specification style document? Otherwise I feel this page completely defeats the purpose of putting this information online for people to learn about something new.--Ottolodr (talk) 15:59, 1 April 2010 (UTC)

I remain concerned about this separation. There's nothing wrong with separating routing protocols from router/routing, but I don't see a clear distinction between "router" and "routing". Unless one simplifies "router" to the point that it is no more than a "box that interconnects networks", I really don't see a clean place to split. Some people use "routing" to express what is properly "forwarding", but the IETF doesn't use the term in that manner.


 * You are correct, it is not clear. The articles existed this way and when I began the cleanup It seemed wise to leave them seperate. The Router article has to of course talk about routing but routing article should be a much more expanded explaination of the topic. In the same way Control Plane is a seperate stub from Router but is also included in router. Howard C. Berkowitz 17:40, 29 June 2007 (UTC)

You could argue that Router / Routing and Routing Protocols all be merged and then use redirect with bookmarking to bring you to the correct portion of the article but it would be overly burdensome to the user of wiki. Remember that you are dealing with the general public. Although the articles are seperate, they are connect with wiki links.

In any event. You can post comment about this here but please rememeber to sign your posts so we know where the comment came from. --<span id="" class="plainlinks" style="color:#002bb8">User: (talk • contribs • count) 15:55, 29 June 2007 (UTC)


 * I'd leave routing protocols separate from router, but merge router/routing, with the more detailed technical discussion of what the router does under control and forwarding planes. To me, it's legitimate for the general public to know that routers connect separate and/or dissimilar networks, but how a router does it is at a more technical detail. The public wants to know aspirin relieves headache and muscle pain, but is probably not interested that it does so by inhibiting cyclooxygenases. Howard C. Berkowitz 17:40, 29 June 2007 (UTC)

As to a different issue, could you comment on the point at which content in an article no longer points to what is considered a stub, but to a more detailed article with considerable detail? I'm thinking here of control and forwarding planes being briefly mentioned in router. I'm substantially enhancing the two "plane" articles with material I don't see elsewhere in Wikipedia.

Obviously, there has to be some balance to avoiding duplication, and perhaps there is some guideline at which point an article is no longer a "stub". Howard C. Berkowitz 17:45, 29 June 2007 (UTC)


 * Stubs are defined here: Stub. Regardless of article size, infomation should not be duplicated to any great extent. Wiki links are suppose to be used. If infomation is duplicated, any editor can just trim the information out.


 * Does the forwarding plane article have enough external links and such to move out of stub? I was going to work on control plane before going back to forwarding plane, and add material, for example, about how routers put routes into the main RIB/routing table. This is not an issue of routing protocols, as different routing protocols (and other mechanisms such as static routes) can present the router with the same route. The control plane logic/RIB manager has to decide which route to use, and, in some cases where the same routing protocol or equivalent mechanism provides multiple routes, tells the forwarding plane to load-balance. I haven't yet written up load-balancing in the Forwarding Plane, but it really applies to both planes. Howard C. Berkowitz 23:32, 30 June 2007 (UTC)


 * I personally, do not do this, because I would rather tell an editor how to do things the proper way to improve your edits. I try to help editors.  Other editors, will just delete the content and they are justified in doing it. I rather let an editor know, so it does not happen in the first place.


 * In additon I am in the Wiki Computer Networking project trying to cleanup a backlog of problems in aritcles. There is alot of interaction that some editors do not realize how one edit is made in article A has ramifications in other articles that need to be fixed.


 * Example: You added an example to the Router article of a Core router. You defined it absolutely correctly. But another editor wrote an article called Core router which discusses what a core router is. His definition also is correct, when you are talking about a Core router on the Internet backbone.  To solve this I had to create a disambiguation article and modify his article to point to your def as well as his, because they are both correct. See: Core router (disambiguation) and Core router


 * Unfortunately, I don't think both are correct. My definition came from RFC 4098, of which I am a coauthor, and, in the team that wrote it, and in the Benchmarking Working Group and the Routing Area, there was considerable discussion. In the RFC, we are principally speaking of provider autonomous systems, which, as much as anything in modern terms, are the Internet backbone. Back in pre-Internet days, when EGP was the exterior routing protocol, it indeed could be assumed there was a single Internet core. That concept, however, has been obsolete for a long time. At best, today's definition of the Internet core is statistical, with the "core" being some number of AS with the greatest traffic according to the CAIDA skitter measurement at http://www.caida.org/analysis/topology/as_core_network/. To the best of my recollection, my own book on Internet routing, Building Service Provider Networks (Wiley, ISBN 0471099228), or other texts such as Geoff Huston's ISP Survival Guide, nor any reasonably recent RFC from the Routing Area of the IETF, uses the term "Internet Core". It isn't used in the next-generation Future Domain Requirements work in the Internet Research Task Force, in which I was a Team B participant.


 * Issue is Howard that you have to keep what was and what is. This is not the IETF and it is not an RFC that gets obsoleted by new information.  There was a core, and the def. of that core and the Core router of the time must stay true to what they were then.  And you can state that this is no longer true now.  But the old data cannot be removed because it was true and this is an ency. that wants all information not just current info. --<span id="" class="plainlinks" style="color:#002bb8">User: (talk • contribs • count) 01:56, 20 July 2007 (UTC)


 * With all due respect, I observe that there is no citation for the earlier definition of "core router". Should't there be one, and from something definitive rather than a trade press article or tutorial? Could the lack of one, perhaps, be, because "Internet core" is a concept that is essentially a misperception by laymen?


 * What happened to "be bold" with material that is wrong? There is a consensus understanding of "core protocols", referring to such things as IP. In point of fact, there never has been a "Internet core" that one can point to on an Internet topological drawing. One might have made that argument, very questionably, for ARPANET or NSFNET, and in the very restricted context of EGP, not BGP.


 * A "core" is a concept much more characteristic of SNA or telephone networks; it runs against the distributed logic of the Internet. If you insist on leaving "core" in, without an authoritative definition, but then want to say the unsourced definition is wrong, I'm afraid the logic of that is something I can't follow. Howard C. Berkowitz 02:14, 20 July 2007 (UTC)


 * The term default-free zone (DFZ), which I'm in the process of correcting, is another way people speak of an "Internet core", but, again, this no longer fits modern practice. It's quite common practice, in a multihomed but stub (i.e., non-transit) autonomous system, for the BGP-speaking router(s) to take "full routes" from the various ISPs to which the AS is multihomed. Especially if there is more than one router connected to the same ISP, a common practice, it will receive more routes that are in the DFZ. Here's the reasoning: when you have two routers connected to a major ISP such as Sprint or Qwest, that provider has a number of customer AS connected to it. The optimal route to those customer AS are important to the ISP itself, but also tells one customer AS which specific router has the best path to the other customer. The "full routes", or properly "full routes plus customer routes", coming to a customer router makes that customer router part of the DFZ, but certainly not part of the current concept of the "skitter core". Howard C. Berkowitz 23:32, 30 June 2007 (UTC)


 * I include all this infomation to let you know that merging articles, redirects, etc. has ramifications to other editor that you may not know are working on other articles. I really dont think these two articles can be currently merged.  I would write the information you would like to write about in the Routing article.  --<span id="" class="plainlinks" style="color:#002bb8">User: (talk • contribs • count) 21:23, 30 June 2007 (UTC)


 * It may be, then, that these ramifications are such that I had best not contribute. Apparently, I don't understand the tag about expert attention being needed. I really don't understand some of the nuanced separations between routing and routers being made here, and if the ramifications are not clearly documented, I can't work with them. Howard C. Berkowitz 02:14, 20 July 2007 (UTC)

Disambiguation
why is there no page named Router(Disambiguation) ? I have one of these, and it does not help me very much in regards to my woodworking needs. —Preceding unsigned comment added by 67.84.191.219 (talk) 23:52, 11 June 2010 (UTC)

Pronunciation
Anyone feel like adding a section on the correct pronunciation of the word, i.e., 'rooter' rather than 'rawter'?
 * Depends where you are from though. --  Thorpe  talk 12:37, 19 October 2005 (UTC)
 * There is an etymologically (and logically) defensible pronunciation, and then there is an American pronunciation. It depends on how fast and loose you want to be with 'right' and 'wrong' I suppose.
 * Actually, there is an etymologically defensible pronunciation, and then there is a pseudo-French pronunciation. The word route (from which this is derived) was borrowed into English from French before the Great Vowel Shift, meaning that the pronunciation rhyming with "out" is the historically correct one. The pronunciation rhyming with "boot" is later, and comes from a desire to "sound more French". —Angr 06:46, 20 December 2006 (UTC)

Someone had added this recently; check the history. I removed it because both are used and pronunciations aren't a typical feature of Wikipedia articles. But please feel free to take this to Wiktionary, where it is both appropriate and needed (currently only 'rawter' is listed). --Rick Sidwell 02:13, 20 October 2005 (UTC)
 * A fair point, but perhaps it will serve a use in allowing some speakers of that fascinating language, American, to see the error of their ways :)


 * If you ever teach routing to Australians, I suggest, from experience, that you avoid rooter. That pronunciation has a very different idiomatic meaning there. Hcberkowitz 02:56, 26 June 2007 (UTC)


 * Transcription seems to be a bit dodgy, can anyone correct that?
 * /'raut@r/ for US, /'ru:t@/ for UK (in SAMPA).
 * Wathiik 15:45, 14 June 2006 (UTC)


 * Actually, pronunciations are common in Wikipedia articles in cases where they might be unclear. But since the /u:/ pronunciation is also common in the U.S., there should be no suggestion that it's strictly a US/UK distinction. —Angr 06:46, 20 December 2006 (UTC)

As the person who convinced the Internet community to switch to the term 'router', from the then-prevalent 'gateway', perhaps I can shed some light.

(I wanted to make the change because 'gateway' also referred to devices that did protocol translations, such as UUCP-TCP mail gateways, so it was confusing to have one term for both. Were I to do it all over again, I probably would have selected 'forwarder', since the confusion between 'routing' (i.e. path selection) and 'forwarding' (i.e. actually handling user traffic) has bedeviled us ever since! But I digress...)

I can tell you that I use both 'rooter' and 'rawter', basically at random. (Then again, I have a UK passport, and I live in the US, so I have both etymological bases covered!) I do the same to the word which is the base, i.e. 'route' - I also pronounce that 'root' and 'rawt', depending on the phase of the moon... Noel (talk) 18:39, 11 August 2007 (UTC)

The IPA pronunciation was once a part of this article. However, someone, or as a part of the cleanup process found it irrelevant; even though I think its handy.

Matty2002 (talk) 11:42, 28 November 2007 (UTC)

I've just tried to re-add the pronunciation, mentioning that both are correct. This was immediately removed by User:Wtshymanski together with an obnoxious comment. Wtschymanski, would you care to either justify why you thing that pronunciation is not tolerable in this article, or else unrevert my contribution? Jec (talk) 18:37, 5 October 2010 (UTC)
 * Wikipedia is not a dictionary and the obnoxious comment, expanded, was "Do these people [who are always correcting your (Jec) pronounciation] even read Wikipedia?". Changing a Wikipedia article to win an argument is pointless. --Wtshymanski (talk) 18:43, 5 October 2010 (UTC)
 * Would you please show me where, in the page you linked, it says that a Wikipedia article cannot contain the pronunciation of a term? Especially in the case of a term that is often wrongly believed to be mis-pronounced? 134.157.168.57 (talk) 01:36, 6 October 2010 (UTC)
 * Read the bit around where WP:NOTDIC where it says encyclopedia articles are about their subjects, not about words. By all means, let's have a long talk page diatribe about this, because of course this article is otherwise so good that it cannot possibly be improved. It's a reliable guide to clulessness on the part of contributors that a discussion about pronunciation, spelling, UK vs. American usage, etc. is longer than any discussion about the factual defects of the article. --Wtshymanski (talk) 03:22, 6 October 2010 (UTC)
 * Sorry, searching for either subject or word in this page yields nothing relevant. Concerning keeping the discussion short -- the discussion would be much shorter if you weren't the one insisting on removing the pronunciation of an often-mispronounced word.  I'm reinstating the pronunciation, sorry. Jec (talk) 22:06, 6 October 2010 (UTC)
 * Oh, I didn't realize you were grepping for keywords instead of reading the policy. Roll on, semantic Web. Still not a dictionary. --Wtshymanski (talk) 00:21, 7 October 2010 (UTC)
 * Wtshymanski, you do realise you're being you bully? Jec (talk) 20:58, 7 October 2010 (UTC)
 * "you're being you bully" is unparsable. What do you mean? --Wtshymanski (talk) 21:55, 7 October 2010 (UTC)
 * Back to the issue, we should include the pronunciation as it is an important part of the topic, which may be difficult for people who hear a vairant term for the first time. The dictionary issue does not apply here because there is far more content than a dictionary entry. Graeme Bartlett (talk) 21:29, 7 October 2010 (UTC)
 * A dictionary is where you look up pronunciations, or "variant". --Wtshymanski (talk) 21:55, 7 October 2010 (UTC)

What is a router?
The article states that "A router is a device that interconnects two or more computer networks". Is that the most correct and clear explanation? As far as I have heard a router can have just two computers connected to it and connect them to an ISP which provide a single IP address. Can a device have just, for example, 3 USB interfaces to connect 3 PCs to an ISP? And is such a device a router? --Lefter 13:15, 3 June 2010 (UTC)
 * Hello Lefter, I think the description is both simple and specific. It makes it very clear that the router deals with networks, not just other computers. An ethernet network HUB can interconnect two or more computers via an ethernet cable. A multiported USB interface is just another form of a HUB. Network sharing can be achived by use of a serial interface cable, a usb cable or a network cable between two computers, only if suitable software in both machines allows it to happen. The key to the description is the word NETWORKS (in the plural sense). A simple Wireless Broadband Router deals with 3 paths, the Telephone line, the wireless broadcasts, and the ethernet hub. It connects the sevices that are requested by devices (computers, Sky TV etc), that are attached to the local (house) network. It ROUTES those requests for access to the Internet Network (remote) computers, or from wireless to ethernet NETWORKS. HUBS unfortunately cannot perform routing functions. They just re-broadcast received PACKETS of information to ALL devices attached to the HUB. I hope this helps you.Francis E Williams (talk) 13:59, 2 December 2010 (UTC)

Advertising?
This article contains a list of companies that either have manufactured or curremtly supply ROUTER devices. This has no place in the information about how the Router works. Should it be deleted?. My answer :- I agree, (with myself), can we have concensus please?Francis E Williams (talk) 14:04, 2 December 2010 (UTC)

Wireless Routers
There is no mention of wireless routers. I wanted to know how they are different from other types of routers. This article doesn't help at all. Twilight Realm 21:49, 3 January 2006 (UTC)

--This depends on how strict you want to be with the term router. The subdomains are likely separated in a wireless system from the subdomain that the wired ports are on, if it has wired ports. To be a true router it must separate the collision domains as well as the subdomains/subnets. You can buy wireless routers that are real routers (e.g., CISCO 1811W) or you could go to BestBuy and get something that is really an "intelligent" switch that really only separates the collision domains and sort of separates the subdomains/subnets. One of the key ways to know if you really have a router versus a switch is the price... general rule of thumb is that if it's less than $200.00 it's not likely a real router. Also, if your wireless device has no wired ports, chances are it is not a router (please note that this is not always the case, but those devices that do not hold true here tend to be professional equipement). Daniel Owens (Dohedo) 18:20, 23 February 2007 (UTC)


 * Wireless routers need no special distinction at all. What you call a "wireless router" is merely a consumer NAT appliance that happens to have a built in wireless network interface built in, or possibly a real router with a wireless network interface attached.  --Afed 14:38, 22 March 2007 (UTC)

Afed is missing the point: there is nothing in the article's description of 'router' that justifies including a very different function -- the wireless access point -- in the product. But the powers that be decided to market it as "wireless router" anyway, paying no attention to the confusion that creates among consumers, a confusion Twilight Realm has quite rightly demonstrated is very real.

The real truth is that what is marketed and sold as a 'wireless router' is really a combination of 'router' and 'wireless access point'. And 'wireless access point' is more than just a "built in wireless network interface": it is not just any interface, but an access point. The "wireless network interface" you have in your computer or on a plug in card does not normally support access point functionality.

That said, I have to admit that I am at a loss to come up with a better term than "wireless router". Then again, I haven't worked at it that hard. The marketeers should have done that.

68.164.80.214 (talk) 22:21, 23 December 2010 (UTC)

Routing transfers to other types of networks
As a retired IT professional who has studied the five volumes and CDs of the MCSE examination in 2002, I have to conclude from some contributors that there exists a misnomer that Routers cannot act as translators or distributors. They are employed typically in the enterprise to do just this. A company in Reading, England uses such a system. It employs both Fibre optic networks within its grounds between two large department stores, a modem telephone link to its deleivery department 5 miles away. An internal NetBEUI (non routable)lan, an internet based TCP(ip) based mailing distribution system using Outlook. All of these were examples of the use of the technology in the real world.Francis E Williams (talk) 18:02, 15 January 2011 (UTC)
 * You've noticed the lack of respect for credentials here, by now. A router ( or a bridge) may translate between different styles of network but that's not essential to its description as a router! Surely that is mentioned somewhere in the 5 volumes of MSCE stuff? --Wtshymanski (talk) 20:09, 15 January 2011 (UTC)
 * The lack of ability for the article to decribe accurately what function(s) the various types of Router perform it what is under debate here. There is no concise single description that allows it to be made plain to the reader what the uses of a router are. It does far more than "connect two networks". It performs many protocol functions within a network. If that concept can be made clear in the lead section it may establish the connections to the further sections in the enlarged article. I cannot comment on what is contained verbatim in the MSCE course, it was a long time ago. I do however, remember it is something that is not being (simply) described in this article. Lack of respect most of us can live with, lack of knowledge some people can live with too.Francis E Williams (talk) 21:04, 15 January 2011 (UTC)

Why?
Why are we confusing routers, bridges and gateways in this article? A router *routes*, it doesn't distribute data any more than a freight train distributes canned goods to your kitchen. A router doesn't necessarily (or even usually) sit between two different protocols. The simplest case of a router could be, for example, all Ethernet around it, and interconnect two TCP/IP networks running on Ethernet. There's no need for protocol changes, but what a router does know is how to get a data packet from network X to network Y, even if it has to hop through A, B and C networks to get there. I don't think the present murky text is any value to the reader and we should revert back to a simpler, correct, explanation. --Wtshymanski (talk) 17:50, 15 January 2011 (UTC)
 * Simples, a router contains a microprocessor which interrogates, (intercepts) the incoming data packet. According to its protocol format and header information it decides (via upgradeable firmware) what to do with it. If it is not suitable for the network that the router is connected to, then it ignores it, it will not forward it. If the packet is suitable to be distributed, (sent), (routed), (dispersed), within the attached network (via the appropriate interface connection) then it will forward it. It will, if neccessary, process and add the correct protocol wrapper for the known type of destination network.


 * It is because the router is so versatile that there is no simple description that is appropriate to allow it to be concise. Using an all TCP ethernet model may seem to be a solution for some readers, others more knowlegeable will require more detailed information. It will be a long and complicated task to reword and clarify this article. Stripping out all the non-router specific information (like hubs, switches, internet backbone and bridges) might clarify it a bit more. Many sections still overlap and contain duplicate material referrals. Adding suitable verifiable content with references to further explanations might help. I have added several more of these references today.Francis E Williams (talk) 00:56, 16 January 2011 (UTC)

Irrevelant references and lack of referencing
Some references are not relevant to the contents of this article. For example: section 1.1 Enterprise routers; "All sizes of routers may be found inside enterprises." reference to "Windows Home Server: Router Setup" web site. Article on Micrsofot website cover home-class routers setup and does not mention enterprise routers. Another two sentences need referencing: The most powerful routers are usually found in ISPs, academic and research facilities. Large businesses may also need more powerful routers to cope with ever increasing demands of intranet data traffic. — Preceding unsigned comment added by Ksylian (talk • contribs) 15:09, 3 March 2011 (UTC)

Routers route, Modems interface different hardware, Firewalls to firewalling
A router only interfaces multiple networks. It is not true that a router is a multiprocessor-based device. Most utilize a microcontroller but they call also be implemented using an FPGA or other hardware. Full-blown PCs have been used exclusively as routers, and a desktop may serve as a SOHO router.

Modems typically convert analog signals to digital signals. What are called "routers" and used to connect to a DSL line contain a modem and a router. A PC can do this also, by using an optional card tp interface with the analog hardware.

Firewalls operate at different layers. Software that does firewalling inside a SOHO router is the same software that does the routing. SOHO routers tend to use Netfilter. The common command used is iptables which can be manipulated by users to do more than what the HTML interface provides by accessing it through via a telnet connection.

Kernel.package (talk) —Preceding undated comment added 22:04, 24 May 2011 (UTC).