Talk:Subnet

Pre-November 2006 Archive

Deleted section

I deleted some stuff earlier from the page (end of Subnets and host count, after 2nd table), which subsequently got added back in (along with some constructive additions). I still think the portion significantly detracts from the article, but rather than end up in an edit war I'll just put my thoughts here and leave it up to someone else to decide.

Firstly the section was surrounded with comment tags, some of which were nested meaning that only part of the text was rendered (about 50% was just a waste of bandwidth). What is left is a rambling rehash of earlier parts of the article, including parts where the author has added '???' to show they aren't sure what is happening and sections highlighted in bold for no apparent reason.

If anyone can see any redeeming information in this portion it would be useful if they could reformat it and move to the relevant sections of the article. Tjpayne 20:19, 24 July 2007 (UTC)

Sorry, either the comment tags didn't show or I missed them. It wasn't clear why the section was deleted en masse, but, when I looked at the previous text, my biggest concern was the continued use of classful terminology. I've been teaching classless methods since CIDR was first developed, and I find that thinking in classful rather than classless, and octets rather than binary & prefix length, are the greatest obstacle to people understanding subnetting. I'd be glad to try to work out some clarification now that I have a better idea what you are trying to do.

Just for reference, when I teach subnetting, I start out completely in binary until I'm certain people see the context of the prefix, and then introduce the /nn notation to denote prefix length. I then show subnetting is a matter of extending prefix bits (or increasing /nn value), while supernetting/aggregation is a matter of collapsing prefix bits (or decreasing /nn value). Only after the students can see clearly what is happened do I introduce dotted decimal notation, describing it as a way of making the binary more human readable. If some of the class understands hexadecimal, or if I'm also teaching IPv6, I show how hexadecimal is another means of human representation.

Only after students understand what the binary address and prefix are doing (including Boolean operations to extract the prefix from an address), and can write them out in dotted decimal, do I introduce classful concepts, deprecating them as obsolete but things that they may encounter. Later, if I am also teaching routing protocols, I'll show all the problems that classful routing causes.

I've done this long enough that I'm comfortable that it is a better way to learn than anything starting with octets. Yes, not all students are comfortable with binary, and if they are not, I drill that first. Sooner or later, they will have to master it. Howard C. Berkowitz 20:34, 24 July 2007 (UTC)

somewhat misleading

"For example, 255.255.255.0 is the network mask for the 192.168.1.0/24 prefix." while this is true, 255.255.255.0 is the network mask for anything that ends in 0. so it's also the mask for 208.33.47.0 a more accurate mask would be 192.168.1.0. applying that mask to anything in the 192.168.1.0/24 range will always yield 192.168.1.0 (so 192.168.1.101 masked with 192.168.1.0 will yield 192.168.1.0 but 208.33.47.0 masked to 192.168.1.0 will yield 208.33.47.0 and thus not be routed with this mask — Preceding unsigned comment added by 74.84.117.50 (talk) 18:35, 18 April 2013 (UTC)

It might help to have an explanation of how the octets add up

I'm not going to make the edit in the article for fear of not being clear enough. :) But I will add it here.

What might help people is seeing how the bits get added up. Everything progresses rather logically when you look at it in binary but it's the translation to decimal that makes everything look so random.

Bit 1	Bit 2	Bit 3	Bit 4	Bit 5	Bit 6	Bit 7	Bit 8
0	0	0	0	0	0	0	0
128	64	32	16	8	4	2	1

With no bits selected, the value is 0.

Bit 1	Bit 2	Bit 3	Bit 4	Bit 5	Bit 6	Bit 7	Bit 8
1	1	1	1	1	1	1	1
128	64	32	16	8	4	2	1

1+2+4+8+16+32+64+128=255

Bit 1	Bit 2	Bit 3	Bit 4	Bit 5	Bit 6	Bit 7	Bit 8
1	1	0	0	0	0	0	0
128	64	32	16	8	4	2	1

128+64=192

This should make the whole idea of borrowing bits here and there more sensible.

--Gmuir 13:40, 17 May 2007 (UTC)

Some changes could be done, and thanks to your message I generally feel more compelled at changing the article a bit or adding an example of some sort. (generally) Logictheo 20:58, 2 June 2007 (UTC)

I'm sure there must be an article on wikipedia about binary numbers. This is an article about what a subnetwork is, not a guide on binary numbers. If explanation is required on this subject it can be linked to.--Crossmr 16:35, 3 June 2007 (UTC)

EMBOH ORA WERO AQ — Preceding unsigned comment added by 114.79.48.34 (talk) 05:32, 15 December 2012 (UTC)

This article needs clarity?

I consider myself fairly technically savvy, but I’m no network admin. I found this article is highly technical and presents more theory rather then explanations. I might suggest it’s a challenging for anyone to understand this who doesn't already have an extensive knoweldge of networking terminology. I found the external link to the about.com article far more effective at describing what subnetworking really is with better examples. --Trode 18:31, 1 December 2005 (UTC)

Sorry about that. I was trying to leave it pretty simple, but I'm sure it can be dumbed down some more. Taking too much away from the article would leave a lot of room for explanation, but simplifying the basics at the start would probably help. Strip away too many of the details and it becomes little more than an overglorified dicdef. Math articles tend to have the same problem. --Gamera2 20:59, 3 December 2005 (UTC)

I agree, Trode. I can't make head or tail of it. Maybe add more explanation 88.111.52.30 15:17, 29 December 2005 (UTC)

This article really needs more clarity. There is, for instance, no clear definition of what the subnet mask is or what role it plays in networking/ TCP/IP. Why are there only certain numbers allowed? How does it work exactly? The article assumes a certain amount of knowledge on behalf of the reader, which is fallacious encyclopedic practice.--Hieronymus 09:04, 20 April 2006 (UTC)

I agree it could be much clearer. I have an EE degree and do a bit of network administrating on the side, but still found this a very difficult read.—The preceding unsigned comment was added by 71.113.121.17 (talk • contribs) 06:51, September 28, 2006.

I also agree that this article needs some rewording and cleanup. Its certainly not very clear even for someone who knows what subnetting is and how it is done.--Crossmr 16:27, 28 November 2006 (UTC)

I looked up "Subnet Masks" as I wanted to know more about the subject. I was redirected to this page, to notice that "Subnet Mask" have an article link here even. Not being too much involved in the Wikipedia community, not knowing the rules, I'm posting here instead of editing the page. Should Subnet Mask be linked in the third paragraph? Johnathon. —The preceding unsigned comment was added by 212.159.122.209 (talk) 12:40, 6 December 2006 (UTC).

i added a "simplify" tag for the reasons stated here—The preceding unsigned comment was added by 165.173.14.76 (talk • contribs) 01:30, December 14, 2006.

If those are your only reasons then it can be removed. The article has been almost completely rewritten since these comments were posted.--Crossmr 02:08, 14 December 2006 (UTC)

Is there a typo under the 'Private subnets' subtitle? When you borrow 2 bits and examine the networks, shouldn't you get 192.168.1.0, .64, .128 and .192 networks only? .0 being all zeros and .192 being all ones taking those out. Then the theory behind calculating the amount of networks you get would match also (2^2-2 = 2).

Cisco equipment allows the use of subnet zero, I'm not sure if other companies hardware allows this or not as well.--Crossmr 23:56, 29 March 2007 (UTC)

That still doesn't address the issue of where the 5th subnet comes from. I'm still learning but after .192 aren't all the bits switched 'on' hence you can't get to .255 without borrowing more bits? Also the theory presented in the preceding section states that by borrowing 2 bits you get 2^2 = 4 networks not 5 (RFC1812) (or 2^2 - 2 which equals 2 not 3 (RFC950)) —The preceding unsigned comment was added by 82.153.36.233 (talk) 17:03, 10 April 2007 (UTC).

It's not a matter of Cisco support; it's a matter of CIDR/VLSM support. The only reason there is a "subnet zero" problem is as an artifact of obsolete classful addressing. If you are relying on the first octet range to tell you the basic subnet mask, you can't tell if 10.0.0.0 is an 8-bit prefix and 24-bit host field, or a 24-bit prefix with an 8 bit host field, or a 28 bit prefix with a 4 bit host field.

If, however, you are given the prefix length, which is always written out for human CIDR compliance, and is always available in classless routing protocols, there is no ambiguity among 10.0.0.0/8, 10.0.0.0/16, 10.0.0.0/24, and 10.0.0.0/28.

The sooner people stop teaching about, or thinking about, classful networks, the easier IP addressing will be. When you think of IP addresses as binary strings, not octets, there will be great understanding. Unfortunately, too many people, affected by obsolete classful addressing, focus on octets as having meaning for the actual address. Octets are merely a way of writing things out in one human-readable form. It's not an accident that IPv6 addresses, with exceptions for IPv4 compatibility, are not just written with the CIDR length prefix, but in hexadecimal. Octets and dotted decimal are major and unneeded sources of confusion. Howard C. Berkowitz 04:40, 14 July 2007 (UTC)

Rewrite

Several people have been expressing concern that this was hard to understand. I've spent the last 90 minutes rewriting this article in, what I hope, an easier to understand method. I included some binary examples. I found that when I learned how to subnet seeing it as binary really made things easy for me to understand and grasp the concept. I've only copy edited it a small amount, so I'm sure it does require some of that. I've kept the links as I didn't have the time right now to go through them and make sure they all apply and I've kept a few of the paragraphs and one of the tables that was there before. I've tried to lay down a foundation of what a subnet mask is, what a network address is and how they work together. Any comments are welcome.--Crossmr 17:47, 28 November 2006 (UTC)

I've re-written the 'Subnets and host count' section to make it (hopefully) clearer and to better represent current practice. I've also removed the section on private addresses as it contained a large number of errors and was largely not relevant to subnetting. Most of the information is already in the IPv4 article - which seems a more sensible place for it (in particular the address ranges used for private IP addresses are not directly relevant to sub-netting). Tjpayne 18:39, 15 April 2007 (UTC)

Image caption

A graphic representation of relationships and source of the various variables representing a chunk of C subnets

"chunk" is an unprofessional term being used here, and that wording is confusing for someone who doesn't already understand subnets, IPv4 and all the terms associated with it. Referring to it as Class C makes it more clear what is being talked about. The other wording:

A graphic representation of the possible lengths of subnets in a class C network

Is a more accurate and concise description of what it is. They are indeed possible lengths, its possible you could have a subnet mask of any of those lengths in a class C network. It will certainly be one of them, but it its possible to be any of them. The image is certainly educational, which is why I didn't remove it, but it needs to be described better.--Crossmr 16:56, 30 November 2006 (UTC)

I agree that my description suffers from substandard wording but it does so for the higher purpose of acuratelly describing the image, to work on a better description i would like to highlight the features of the image that need to be reflected in it's description.

1. it represents the last of the 4 octet blocks forming ALL of the 32 possible CIDR blocks , all 4 blocks being variations of a base data which is represented here , half of the C class (hence chunk and C)
2. it depicts the way one value can be extracted from another and where do all the values come by color code and formula's in the header (hence relation and source)

i have nothing against anyone crafting a description to both reflect these and be fluent--Mancini 17:28, 30 November 2006 (UTC)

edit id 91277922 abusive edit

In relation to the http://en.wikipedia.org/w/index.php?title=Subnetwork&diff=91277922&oldid=91271869 edit , i agree with moving the image but otherwise please do not edit my edits anymore , concerning your stated reason for censoring my last edit the data was allready stated , which is false , not once does network address translation is mentioned.

I am in the process to rewrite this article in concordance with the Wikipedia:WikiProject_Computer_networking guidelines , and i was pointing out the history of subnets (reasons why it exists , alternatives) data that need to be in the header

Please comment on the talk page next time you have a issue about my edits not censor them.--Mancini 15:11, 1 December 2006 (UTC)

Please see WP:OWN if you don't want your edits edited by other editors, do not edit here. The text you added was already covered further down in the article and it was unnecessary to repeat it in the beginning of the article. I've reinserted the part about the IPv4 shortage. NAT wasn't mentioned down below, but NAT has nothing to do with Subnetworking. However I'd already covered the concept of using fewer addresses by using subnetworks below. I've reinstated that piece as a bit of a summary. If someone doesn't understand subnetworking at all, giving a very brief and technical overview about borrowing bits isn't going to help them. As far as the guidelines go, they're guidelines and not policy regarding computer related topics.--Crossmr 15:24, 1 December 2006 (UTC)

I do not have anything against the others , i resent your bad-faith censoring of my edits , i will present new content for voting on this page from now on and seek technical people to override your questionable expertise on the subject.--Mancini 16:24, 1 December 2006 (UTC)

I didn't "censor" your edits in bad faith. I resent your bad faith assumption that I was doing it in bad faith. As far as technical expertise goes, if you've got a question about the way I've described something, feel free to ask it. You might also want to read WP:NPA comment on content not the contributors.--Crossmr 16:31, 1 December 2006 (UTC)

Whatever reason your edit was non-constructive , Intentionally making non-constructive edits to Wikipedia will result in a block or permanent ban and instead of debating about it , i will do as i said and provide new content for voting on the talk page , i am not going to argue with someone that obviously does not even know that in computing a "chunked structure" represents recurrent chunks of slightly variable data--Mancini 16:41, 1 December 2006 (UTC)

and attempting to own your contributions or making personal attacks will result in the same thing. My edit was constructive in repositioning an image and removing content that was covered further down in the article. You properly raised the point that perhaps not all of it should be removed and it was put back in. As far as chunks perhaps you should tell google, because "chunked structure" isn't a very widely used term.[1] which seems to mostly have to do with a specific file system type which have nothing to do with subnetworking, and an occasional linguistic reference. Using chunk in that context doesn't properly describe what that is--Crossmr 16:52, 1 December 2006 (UTC)

I do not know what the problem is , but you could have said you reposted relevant data , i assummed bad-faith because we should work toghether , your actions point against.Do i really have to give google tips ? try "chunk structure" , it is widely used to describe data , and quite so in networking relating to packet structure , your description of chunk as an unprofessional term in the context is clearly without foundation.--Mancini 17:05, 1 December 2006 (UTC)

Packet structure has nothing to do with the process of subnetting a network. The term is not used to describe subnets of a network unless its being used as slang. [2] [3] I see no evidence that the term is commonly used in regards to subnetting or networks in general, outside of speaking about the content of packets.--Crossmr 17:10, 1 December 2006 (UTC)

Merge Suggestion

I don't see the point of a merge here unless its to make the CIDR article huge. I could only see it being merged to have someone suggest that it be split up because both pieces are quite lengthy.--Crossmr 02:31, 4 December 2006 (UTC)

I also dissagree with the merge, quite the opposite should be done and move all info about the subnetting technique to it's own article and improve this one to describe clearly and in detail the subnet/subnetwork concept and it's operation.--Mancini 14:02, 4 December 2006 (UTC)

Wrong table?

Currently, the article has the following table:

Class	Leading bits	Start	End	Default Subnet Mask in dotted decimal	CIDR notation
A	0	0-127.x.x.x	126.255.255.255	255.0.0.0	/8
B	10	128-191.x.x.x	191.255.255.255	255.255.0.0	/16
C	110	192-223.x.x.x	223.255.255.255	255.255.255.0	/24
D	1110	224-239.x.x.x	239.255.255.255
E	1111	240-255.x.x.x	255.255.255.0

Shouldn't it be 0-126.x.x.x in the field Start | A? --Abdull 17:23, 10 April 2007 (UTC)

i would argue that although the 127 network only has device-local relevance and therefore cannot be assigned to a network, it is still technically a class A allocation. The table is still wrong, but for a different reason :) Tjpayne 18:52, 15 April 2007 (UTC)

For consistency I've updated this table to use the same one as IPv4 subnetting reference Tjpayne 14:51, 18 April 2007 (UTC)

References

I'm a little concerned that third-party Cisco certification preparation guides are being listed, here and in other networking articles, as references seemingly at the same level of authority as RFCs and even vendor-independent texts. From personal experience both as a Cisco instructor and in the preparation of certification preparation material, these sometimes differ from such things as the primary standard, because the author(s) understand that the certification tests are usually based on Cisco courseware, which sometimes has obsolete or proprietary material. This is not to say that the certification guides aren't accurate for the purpose, but they aren't definitive for the field.

When RFCs from the IETF address the subject, especially if they are Standards Track, they are authoritative. Now, with all proper disclaimers, I have written vendor-independent networking books, such as Designing Addresing Architectures for Routing and Switching from MacMillan. Other authors also have written books that were not strictly written for Cisco purposes. I have also written certification study aids at certificationzone.com, and I would consider the vendor-independent material more definitive. In some of the study aids for which I was responsible, the authors would point out there is a "right way, a wrong way, and a Cisco way." Bottom line: if an RFC speaks to a subject, it should be considered definitive unless there is overwhelming evidence to the contrary, or it is superceded by another RFC. Howard C. Berkowitz 13:20, 19 July 2007 (UTC)

Do subnet masks in non-standard format exist?

It would be useful to find on this page something defining the format of a subnet mask. What I've not found is: subnet mask can be only in the format 1...10...0 or can zeros be also inside 1s part?

Do you mean 11011...etc? no. Subnet masks are 1s followed by 0s in binary format.--Crossmr 12:51, 23 July 2007 (UTC)

A subnet mask is a 32-bit binary string, which has been clarified by CIDR RFCs to have the requirement that all one bits must be contiguous from the most significant position on the left. In other words, you can have something that would be 11111111 00000000 00000000 00000000, but not 11111111 00001111 00000000 00000000. There are two ways to represent, in a human-readable way, IPv4 subnet masks: dotted decimal similar to IP addresses (e.g., 255.0.0.0), and prefix length indicating the number of 1 bits (e.g., /8).

IPv6, however, uses prefix length only, and represents the IP addreses in hexadecimal. There is one exception: IPv4 addresses embedded in IPv6 may have the IPv4 part in dotted decimal. Howard C. Berkowitz 13:56, 23 July 2007 (UTC)

IPv4 Classes section needs tidying

There is some stray text in italics above the first table and the text (about loopback, etc.) below the table doesn't make sense.

New merge suggestion

I propose that IPv4 subnetting reference be merged into this. There's redundant information, and the other article seems very incomplete. —Preceding unsigned comment added by Reinderien (talk • contribs) 02:36, 11 June 2008 (UTC)

I disagree. If any merge is to be done, sections of this page should be merged into IPv4 subnetting reference. The title is more descriptive, and the information is more detailed than on this page.NetworkFloridaDOTcom (talk) 15:26, 24 October 2008 (UTC)

I disagree. This page is strictly about IPv4 addressing, whereas the Subnet article should discuss subnetting in all versions of IP. I will remove the template, since this has been posted for some time. Kbrose (talk) 21:35, 31 October 2008 (UTC)

Please check this edit/deletion

Are the changes made in this 31 Dec 2008 edit http://en.wikipedia.org/w/index.php?title=Subnetwork&diff=261020601&oldid=260259765 really an improvement? Seems to be a fair bit of material deleted and the opening sentence seems worse. However I do not have the knowledge to know for sure. Nurg (talk) 23:10, 14 January 2009 (UTC)

Move the example to be near the top

Please move the example with the cable modem near the bottom of the article to be near the top, its the clearest explanation of the concept across the whole article. If a diagram illustrating the relationships could be added that would really make it clear! —Preceding unsigned comment added by 65.91.42.118 (talk) 14:23, 7 May 2009 (UTC)

Should 255.255.255.0 lead here?

From the standpoint of a Wikipedia reader (and not an editor), it would make sense that one of the more common subnet mask addresses lead here. For example, I typed in 192.168.1.1 and it lead to the localhost article. Should 255.255.255.0 similarly be redirected here? I would do it myself but I have no idea how to do it. Thanks for reading —Preceding unsigned comment added by 207.255.210.182 (talk) 14:24, 10 May 2009 (UTC)

I see the logic, and it seams like a sensible idea. But then the question comes up "where do you see 255.255.255.0 and not be told it is a subnet mask?" As for the example you gave, I could see 127.0.0.1 redirecting to localhost, but shouldn't 192.168.1.1 either to default gateway or more likely the article focusing on private addresses? But I'm no editor, a poor writer, an just use Wikipedia anon. 97.122.165.186 (talk) 22:29, 31 July 2009 (UTC)

Clarify AND operations

"When Foo sends data to example.com at 208.77.188.166, the router performs a logical AND of the destination example.com address with the subnet mask. It also performs a logical AND of the origin address (17.76.99.1) and recognizes that these two results are different, and therefore sends the data over the Internet, via the subnet's default gateway." "It also performs a logical AND of the origin address (17.76.99.1)" It performs a logical AND with the origin address and what? AND operations involve two cases. It would also help to have the binary code stacked on top of eachother so you do the AND operation for yourself in your head more easily.173.70.15.67 (talk) 14:24, 13 September 2009 (UTC)flat9@hotmail.com

---

to point out, it's not a LOGICAL AND, its a BITWISE AND. logical and is eg. if(1>2 && 2<3) while bitwise and would be eg. 3 & 2 = 2 (tus) — Preceding unsigned comment added by 74.84.117.50 (talk) 18:15, 18 April 2013 (UTC)

Implementation section

I came to the page looking for answers and information for implementing subnets, this page didn't really help. While it may be very useful in some categories, and KUDOS to those editing it, it is missing a section on how to implement subnets. Sure subnets are useful, here is what they are useful for, here are some changes occurring in the networking space, etc. Where is the implementation section? How to use a subnet. Think of your audience as technically inclined individuals who are trying to set up home networks, or office networks that don't feel like calling the cable guy. eximo (talk) 22:22, 12 October 2009 (UTC)

WP is not a how-to manual. See WP:NOTHOWTO Kbrose (talk) 02:10, 13 October 2009 (UTC)

C-class reassessment

per request. Not suitably reference (only 4 at the time of writing!). Airplaneman ^talk 05:14, 14 November 2009 (UTC)

Perhaps this editor should actually read the article and not only look in one place. Article contains enough references, in addition to inline RFC references to support the subject, given that it is a narrow topic. Reverted action.Kbrose (talk) 14:48, 14 November 2009 (UTC)

Usable hosts

Whoever wrote up the total useable hosts bit is smoking crack. Someone please fix this so I dont have to find some other random article on subnet masks to show people who dont know what subnetting is. I'll let a wiki savvy person do that, but I wanted to point out this serious fundimental error. --64.69.87.237 (talk) 07:06, 7 January 2010 (UTC)

I stared for a while and could not see a pattern as to what those numbers meant and so went ahead with changing them to numbers that make sense to me. :-)

I came to this article as I'm trying to understand if host 0 (all zeroes) is reserved. I know all zeroes for the network portion of an address is reserved to mean "this" network and all ones of the host portion is reserved for "all hosts on this network." The default gateway has usually been put at host 1 and people avoid zero. RFC 1878 says "Host all zeroes and all ones excluded. (Obsolete)." RFC 943 dances around it with:

"In certain contexts, it is useful to have fixed addresses

with functional significance rather than as identifiers of

specific hosts. When such usage is called for, the address

zero is to be interpreted as meaning "this", as in "this

network". The address of all ones are to be interpreted as

meaning "all", as in "all hosts". For example, the address

128.9.255.255 could be interpreted as meaning all hosts on

the network 128.9. Or, the address 0.0.0.37 could be

interpreted as meaning host 37 on this network."

Unfortunately, that did not talk about 128.9.0.0 which is what I'm asking about. --Marc Kupper|talk 10:18, 3 March 2010 (UTC)

Why does subnetting help efficient use of adress space?

To me it sort of seems to be the other way around, since each device needs one adress in each subnett it is a part of. So subnetting means that routers needs more adresses. In addition each subnett must have a power of two adresses, so some adresses will be unused. It would seem to me to be more efficient, in terms of adress use, to not subnett at all, although this would make routing a nightmare. So in conclusion, isn't the reason for subnetting facilitation efficient routing, rather than saving adresses? 213.161.190.227 (talk) 05:41, 28 April 2010 (UTC)

Many LANs have far fewer nodes than the number of available addresses in a full network. Rather than burning up a network number for each LAN and wasting many addresses we subnet. For example, for a while I had four computers at home that were on the Internet. I ordered an eight address subnet from my ISP which gave me five usable address. I can't use address 0, address 1 was the default gateway and my ISP's router, addresses 2 to 6 were available for my use, and 7 was the broadcast. This was much cheaper than ordering a full class-C network. IPv4 only has two million class C networks meaning there's clearly no way we can provide one to every home user.

An example of a business use is where I have a 128 address subnet in a data center. In that case the data center is itself has a 16,384 address subnet of a class A address and the data center subnets that out for their customers as needed. The Internet saves on addresses as they did not need to allocate a full class B or A space to the data center. Technically the data center could implement the LAN as a single flat network with 16K addresses. However, that means that broadcast messages and ARP resolution would hit all machines on the LAN. Assuming the LAN is Ethernet the data center would end up needing to install a 16K port Ethernet switch to avoid flooding the LAN with traffic. A flat address space also complicates management of the IP address space for the data center. The solution to all of this is that the data center subnets and each customer manages their own network.

Customers also want to define their own routing, bandwidth shaping, and firewall rules and it's far easier to do this if they have their own router and firewall meaning they will be managing their own (sub)networks. --Marc Kupper|talk 04:13, 29 April 2010 (UTC)

There really are a lot of misconceptions about the utility of subnetting in the enterprise. The article used to assert that it has all kinds of efficiency gains without really analyzing any situation or citing references to particular deployments. I tried to eliminate these assertions by suggesting only the possibility using the phrase 'may be' often. This was reverted recently, but I felt compelled to reestablish my version as a better more balanced presentation.

In historical hind-sight, one wonders what would have happened if one could just get another address allocation easily and connect a sub-department with its own router. This would likely be the preferred method as it assures the end-to-end principle everywhere. But with private address spaces, NAT, etc. to combat the shrinkage of allocation pools this was unfeasible, and it stands to argue that the only benefit of subnetting in the enterprise is the conservation of global address space and the benefit of keeping global routing tables smaller. This is exactly what CIDR was supposed to accomplish and did until assured exhaustion. In most other aspects in the enterprise subnetting creates other inefficiencies in exchange for some gains. Without going into detailed analysis, I would bet that every one of the perceived benefits can be achieved by other means. Subnetting and NATing became a sport, network managers engaged in, and the schemes can be amazingly complicated, but it is hard to find reliable references for an encyclopedia to use to justify reporting efficiencies. Kbrose (talk) 20:20, 16 September 2010 (UTC)

Of course I was not trying to undo your edits - just trying to make what was already there easier to read. Thanks for bringing the error to my attention. I've posted a revised version - added back in a may and removed any mention of address space efficiency - that is disputed here and not supported in the body of the article. It shouldn't have been in there in the first place. --Kvng (talk) 23:36, 16 September 2010 (UTC)

Routing prefix

This was referred to as network number, routing prefix and network prefix. For consistency and clarity, I've changed all to routing prefix (and prefix for short) because that's what was predominantly used in the article. We probably want to go with the term used predominantly in the literature. I'm thinking that's network number but I'd like to solicit opinitions and citations on this. --Kvng (talk) 21:15, 4 October 2010 (UTC)

This sentence appears wrong to me: "The recommended allocation for an IPv6 customer site is an address space with an 80-bit (/48) prefix". I think it should be 80 bit suffix. The referenced page (ref 16 - getipv6.info) says: "No subnets will use prefixes longer than /64. " which to me says the maximum prefix length is 64 bits. — Preceding unsigned comment added by 209.228.188.200 (talk) 11:37, 5 March 2012 (UTC)

Very confusing article

This article is way too confusing and haphazard. A far better and to-the-point explanation about subnets and masks and what those numbers mean is here: http://www.iplocation.net/tools/netmask.php — Preceding unsigned comment added by 24.22.219.163 (talk) 12:01, 23 July 2012 (UTC)