IP adress for small network that needs internet access
Nevyn
ok....got one computer....may get another soon.....i'll want to make a
network of the 2 of them....mostly cos i like to fiddle.....if i set up my
SuSE 6 box as if i have the network already as far as possible......what
would be a good IP adress range to give the computer considering i go on the
internet with it....or does ot not matter?
*************************************
** "yurtta sula cihanda sula" **
*************************************
David Means
are only a couple of choices:
1. PPP over serial line. Your ISP (dynamically) assigns the IP address
for this connection, so you probably aren't even aware of it. On the
ethernet side, (which currently goes nowhere), use one of the RFC1597
(now RFC1918)-specified private address blocks. [ For home use, I
recommend 192.168.n.{1-254}. ]
2. DSL attached to an ethernet port. Your ISP gives you (static) address
which you use for that ethernet port, or they assign it automatically. In
either case, you use a private-address-block number for the 2nd ethernet
port that is attached to your local net.
Nevyn <ne...@nevynxxx.freeserve.co.uk> wrote in message
news:7krner$6ir$1...@news5.svr.pol.co.uk...
Nevyn
change
addresses for your local network using the existing box as the gateway for
the other box(es).
HTH
Michael Black
Nevyn wrote in message <7krner$6ir$1...@news5.svr.pol.co.uk>...
wclar...@my-deja.com
"David Means" <me...@nospam.digitalelephant.org> wrote:
> [...] use one of the RFC1597
> (now RFC1918)-specified private address blocks. [ For home use, I
> recommend 192.168.n.{1-254}. ]
I have often wondered about this. Why is it that so many people suggest
using 192.168/16 [actually, most recommendations are for 192.168.x/24]?
I always suggest using 10/8 [10.0.0.0-10.255.255.255 for those
unfamiliar with the "/" notation]. On a private network, there is no
reason to conserve IP space, and I prefer the extra room the 10/8
network provides for custom naming.
One interesting scheme I saw worked something like this:
10.a/16 - Geographic location A
10.b/16 - Geographic location B
10.c/16 - Geographic location C
etc...
10.a.1/24 - All devices of type "1" [routers] at location A.
10.a.2/24 - All devices of type "2" [eg, DNS] at location A.
10.a.3/24 - All devices of type "3" [eg, DHCP] at location A.
etc..
10.a.x.1 - router for network X in location A
10.a.x.2 - DNS for network X in location A
10.a.x.3 - DHCP for network X in location A
...
10.a.x.100 - host on network X in location A
etc...
Then, you set up aliases such as
10.a.1.x [router for X at A] for 10.a.x.1
So that you can do nifty things like:
`ping 10.a.1.255`
..and see that all your routers are up at a specific location.
Very cool stuff.
For learning about networking, subnetting, and aliasing, it helps to
have a lot of "space" to play in. Hence, why I always suggest using
10/8.
--
Bill Clark
Systems Architect
ISP Channel
http://neighborhood.ispchannel.com/
Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
bo...@nospamcyberdude.com
machines
and have no use for more than class C ip addresses.
--
Come Visit Our Website
http://www.freeyellow.com/members/creative-services
Please Visit Our Sponsers (We get paid per
visit)
wclar...@my-deja.com
bo...@NOSPAMcyberdude.com wrote:
> Most people are only working with a hand full of
> machines and have no use for more than class C ip addresses.
That can't be it, because it's no harder to set up a 10.0.n/24 than it
is to set up a 192.168.n/24. There's definitely preferential treatment
given to the subnets of 192.168/16. It has to stem from *something*,
since random chance would favor 10/8 [since it's more useful].
Todd Knarr
> That can't be it, because it's no harder to set up a 10.0.n/24 than it
> is to set up a 192.168.n/24. There's definitely preferential treatment
> given to the subnets of 192.168/16. It has to stem from *something*,
> since random chance would favor 10/8 [since it's more useful].
Laziness? New users ask someone who's already set up a real IP network
for advice. The people they ask probably got a class C network number,
and they use a class C private network number so they simply can take
their configuration, change the network numbers and use it pretty much
verbatim. This leads to more people knowing how to set up a class C
private network, who pass this on to others, and so on.
For myself, it's philosophy. I use the smallest class of private network
number likely to handle the needs of the network. Home networks aren't
likely to exceed 254 hosts, so class C fits. For a company I'd pick the
10 network and do real subnetting, since sooner or later they're going
to need the address space ( assuming they don't go out of business first ).
The class B private networks seem to be pretty much ignored.
--
Collin was right. Never give a virus a missile launcher.
-- Erk, Reality Check #8
wclar...@my-deja.com
Todd Knarr <tkn...@xmission.com> wrote:
> Laziness? New users ask someone who's already set up a real IP network
> for advice. The people they ask probably got a class C network number,
> and they use a class C private network number so they simply can take
> their configuration, change the network numbers and use it pretty much
> verbatim. This leads to more people knowing how to set up a class C
> private network, who pass this on to others, and so on.
But that only works for historical reasons -- there is no such thing as
a class C network any longer [they're now part of the /24 networks].
Some older ethernet cards cared what network class you were using, but
those haven't been on the market for many years. For most people, there
is absolutely no difference whatsoever between using 10.x.y/24 versus
192.168.z/24. Everything will "carry over" in exactly the same manner.
> For myself, it's philosophy. I use the smallest class of private
> network number likely to handle the needs of the network.
Networks aren't in "classes" any longer, and haven't been for roughly a
decade. 10.x.y/24 is the same as 192.168.z/24 in every conceivable way,
except that addresses from the first net are capable of being part of
much larger super-nets than those from the second.
> Home networks aren't likely to exceed 254 hosts, so class C fits.
Here you really mean /24 [not class C], and that doesn't preclude using
a /24 subnet of network 10.
> For a company I'd pick the 10 network and do real subnetting, since
> sooner or later they're going to need the address space ( assuming
> they don't go out of business first ).
I'd love to see the company that "needs" 16.7 million IP addresses. /8
networks are used so that detailed subnetting can occur, and for no
other reason. My problem is that most people limit themselves to a /24
subnet of a /16 network, which doesn't leave a whole lot of room for
fancy subnetting. Granted, a lot of people will never care about
learning about [ie playing around with] subnetting. However, that still
doesn't explain why 192.168.z/24 is used almost *exclusively*,
especially considering there are no benefits at all to using such a
network over something from 10/8 [or even 172.16/12].
> The class B private networks seem to be pretty much ignored.
That's an excellent point. 172.16/12 [which is what you mean -- "class
B" no longer exists] *is* almost completely ignored nowadays. A while
back, I posted about this as well. Now that you've brought it up, I'll
ask again: Does *anybody* use the 172.16/12 private network? ANYONE?
This whole phenomenon of using 192.168/16 almost exclusively, using 10/8
on rare occasions [most new companies, at least those I've dealt with,
use 192.168/16 and *not* 10/8], and _NEVER_ using 172.16/12 has me
completely baffled. This is _not_ how it was five or six years ago,
when .edu made up the bulk of the internet. I haven't seen a single
172.16/12 address since I stopped working at a University.
Of course, none of this is really a problem [the users of
poorly-designed networks are the ones who lose out -- they are, after
all, *private* networks, so it doesn't affect me in the slightest :)].
Nonetheless, it disturbs me that I can't find a good *reason* for it.
Michael Fuhr
> In article <7l19do$6bq$3...@news.xmission.com>,
> Todd Knarr <tkn...@xmission.com> wrote:
>
> > The class B private networks seem to be pretty much ignored.
>
> That's an excellent point. 172.16/12 [which is what you mean -- "class
> B" no longer exists] *is* almost completely ignored nowadays. A while
> back, I posted about this as well. Now that you've brought it up, I'll
> ask again: Does *anybody* use the 172.16/12 private network? ANYONE?
I know of at least four large companies that use 172.16/12. At least
three of those also use 10/8, 192.168/16, and several public /16s.
It's a real joy when those companies start connecting their networks
and you end up using a lot of NAT to avoid conflicts with the private
addresses.
--
Michael Fuhr
http://www.fuhr.org/~mfuhr/
Todd Knarr
> Networks aren't in "classes" any longer, and haven't been for roughly a
> decade. 10.x.y/24 is the same as 192.168.z/24 in every conceivable way,
> except that addresses from the first net are capable of being part of
> much larger super-nets than those from the second.
True, but a lot of people still have addresses that fit into the old
class A/B/C system. I generally use "class C" to mean "A /24 network
in the range originally assigned to class C network numbers". Ie. the
192.168.171.x network would be class C, 10.56.143.x would not be.
I tend to keep subnetting to byte boundaries a certain amount of the time,
since that gives me IP addresses where the subnet parts can be read straight
out of the dotted-quad form easily. This does lead to the "class" terms
being useful, since they correspond directly to where the boundary lies.
> I'd love to see the company that "needs" 16.7 million IP addresses. /8
Leading octet 10.
Second octet indicating physical building, 1-3.
Third octet indicating segment within the building.
Fourth octet indicating host on the segment.
Works well, since logical departments coincide well with physical
network segments. Requires enough room to subnet on byte boundaries.
Note that in this case "need" is driven more by the need for convenience
for the netadmins than anything else.
My rule is to use class C ( /24 ) on networks that won't need subnetting
for routing, class A ( /8 ) or something else largish on networks that
do ( or conceivably may in the future ) need subnetting. Home networks
typically don't need subnetting, and if they do they're typically owned
by people who know what subnetting is and why they're going to need it.
> That's an excellent point. 172.16/12 [which is what you mean -- "class
> B" no longer exists] *is* almost completely ignored nowadays. A while
> back, I posted about this as well. Now that you've brought it up, I'll
> ask again: Does *anybody* use the 172.16/12 private network? ANYONE?
I don't. If it's big enough to need more than a /24 I usually go straight
to the 10/8 network and be done with it.
> Of course, none of this is really a problem [the users of
> poorly-designed networks are the ones who lose out -- they are, after
> all, *private* networks, so it doesn't affect me in the slightest :)].
> Nonetheless, it disturbs me that I can't find a good *reason* for it.
Part of it might be the people who admin the networks. I know that at
the company I word for the IP address allocation was done by people who
were not familiar with the Internet ( for example, almost all of our
internal hosts are using IP addresses assigned to real machines on the
Internet, and yes we are connected which leads to some interesting
situations ( for painful values of interesting ) ), the new sysadmin
for most of the machines I need to deal with has little Unix and almost
no networking experience and there's serious management pressure to
_not_ go to the "unneccesary" effort and expense of renumbering over to
a real 10/8 private network with proper subnets. I don't have to admin
this mess, so I just keep reminding myself "Not My Problem".
Steve
networking within the corporate infrastructure. 176 for mid-tier and
192 for lower-tier.
wclar...@my-deja.com
Todd Knarr <tkn...@xmission.com> wrote:
> wclar...@my-deja.com wrote:
> > Networks aren't in "classes" any longer, and haven't been for
> > roughly a decade. [...]
> True, but a lot of people still have addresses that fit into the old
> class A/B/C system. I generally use "class C" to mean "A /24 network
> in the range originally assigned to class C network numbers". Ie. the
> 192.168.171.x network would be class C, 10.56.143.x would not be.
I guess it's partly a sign of how long a person has been doing this type
of thing. Just to be clear, my insistence over pointing out the correct
[ie modern] terminology is more to remind myself than anything else. I
figure if I can write "classes no longer exist" enough times, I may stop
using the terminology myself. :)
> > I'd love to see the company that "needs" 16.7 million IP addresses.
> Leading octet 10.
> Second octet indicating physical building, 1-3.
> Third octet indicating segment within the building.
> Fourth octet indicating host on the segment.
Yep. That [among other examples of sensible use of addressing] is why I
never understand use of 192.168/16 private addresses [unless necessary
for some reason]. They just aren't as *elegant*.
> [...]
> Home networks typically don't need subnetting, and if they do they're
> typically owned by people who know what subnetting is and why they're
> going to need it.
You'd be suprised. I've known a great many people that "should" know
what subnetting is [considering they've built large networks at some of
the places I've worked], but still have an irrational aversion to 10/8.
One place -- I kid you not -- went through two different renumberings,
from 192.168.x/24 to 192.168/16 before finally settling on 10/8. Did I
mention that I'd suggested 10/8 from the beginning, but it was turned
down -- I also kid you not -- because we didn't need to "waste" that
many IP addresses? Grr.
> > Does *anybody* use the 172.16/12 private network? ANYONE?
> I don't. If it's big enough to need more than a /24 I usually go
> straight to the 10/8 network and be done with it.
I suppose that makes sense. To be perfectly honest, the only times I've
ever actually seen it in use have been inside multi-layered networks [a
private segment behind a private segment behind a private segment
connected to the internet]. The outermost layer used 10/8 [and had an
addressing scheme similar to what you described above]. Some
departments had their *own* private segments within that [they wanted a
different addressing scheme based on device types, like I described in
my original post], and a few had yet another layer. In order to keep
things as consistent as possible throughout the whole organization
[mostly for the sake of the admins' sanity], and to avoid multiple hosts
having the same address [even though routing and NAT should've --
theoretically -- prevented confusion], different private networks were
used.
> Part of it might be the people who admin the networks. I know that at
> the company I word for the IP address allocation was done by people
> who were not familiar with the Internet ( for example, almost all of
> our internal hosts are using IP addresses assigned to real machines on
> the Internet, and yes we are connected which leads to some interesting
> situations ( for painful values of interesting ) ), the new sysadmin
> for most of the machines I need to deal with has little Unix and
> almost no networking experience and there's serious management
> pressure to _not_ go to the "unneccesary" effort and expense of
> renumbering over to a real 10/8 private network with proper subnets. I
> don't have to admin this mess, so I just keep reminding myself "Not My
> Problem".
I feel your pain.
Villy Kruse
>That's an excellent point. 172.16/12 [which is what you mean -- "class
>B" no longer exists] *is* almost completely ignored nowadays. A while
>back, I posted about this as well. Now that you've brought it up, I'll
>ask again: Does *anybody* use the 172.16/12 private network? ANYONE?
I confess. We use 172.16.xx.yy/26 with possible interconnecting many of
these small networks in the future. Why this was chosen rather than in
10.xxx or 192.xx I don't know, though. Maby the other networks we
occasionally need to connect to are in the 10.xxx or 192.xxx networks,
all unoffical non-Internet network numbers.
Villy
Todd Knarr
> [ie modern] terminology is more to remind myself than anything else. I
> figure if I can write "classes no longer exist" enough times, I may stop
> using the terminology myself. :)
When IPv6 starts to come into common use, the "class" terminology will
go away fast. It just doesn't work for colon-hex notation. One reason
I still use it is that there's still tools that require byte-aligned
subnets and understand the old A/B/C ranges, or that automatically
figure netmasks and such if you feed them A/B/C addresses but require
extra configuration if you use CIDR properly.
> Yep. That [among other examples of sensible use of addressing] is why I
> never understand use of 192.168/16 private addresses [unless necessary
> for some reason]. They just aren't as *elegant*.
LIS, depends on the network. At home I've got 5 CPUs, 2 print servers.
The print servers go on one /28, the 4 working CPUs go on a second, the
firewall/gateway machine goes on a third all by itself. I'm never even
going to fill one subnet, let alone all 14 subnets, so I go with the
smallest network type on the principle of the smaller the range of addreses
I use the less likely I am to ever conflict with anyone else. I'd also
pick multiple 192.168.x/24s if I was dealing with a lot of small,
independent networks that needed to connect to each other ( eg. wiring
several dorm rooms together, where each person wants to keep control
of their own room's computers ).
I guess it's "smallest address range that fits the entity involved",
distinguishing between a collection of entities that need to talk to
each other vs. a single entity large enough to need a logical division
into parts to retain admin sanity.
Bob
On Thu, 24 Jun 1999 wclar...@my-deja.com wrote:
> Date: Thu, 24 Jun 1999 21:56:50 GMT
> From: wclar...@my-deja.com
> Newsgroups: comp.os.linux.networking
> Subject: 192.168/16 vs. 10/8
>
> In article <7krr42$il$0...@216.39.144.210>,
> "David Means" <me...@nospam.digitalelephant.org> wrote:
> > [...] use one of the RFC1597
> > (now RFC1918)-specified private address blocks. [ For home use, I
> > recommend 192.168.n.{1-254}. ]
>
> I have often wondered about this. Why is it that so many people suggest
> using 192.168/16 [actually, most recommendations are for 192.168.x/24]?
My guess is that the average "home network only has 1 location and rarely
more than 254 devices that need connecting to the network.
For learning about networking and subnetting I personally suggest a couple
of good FAQ's including the IP-Subnetworking mini howto.
> --
> Bill Clark
> Systems Architect
> ISP Channel
> http://neighborhood.ispchannel.com/
>
>
> Sent via Deja.com http://www.deja.com/
> Share what you know. Learn what you don't.
>
>
Bob PHILLIPS
Director/System Administrator
b...@norcom.net.au | ISP to the nor'west of Western Australia
| http://www.norcom.net.au
Yes, I am on the interthingy | If it aint broke, fix it, then it will be
--------------------------------------------------------------------------
dotnet dotau Pty Ltd PO Box 2762 SOUTH HEDLAND WA 6722 AUSTRALIA
--------------------------------------------------------------------------
James Knott
>Some older ethernet cards cared what network class you were using, but
>those haven't been on the market for many years.
Since when was any NIC sensitive to network class or even TCP/IP???
Last time I checked, a NIC simply sent a packet from one MAC address
to another, with no consideration for higher level protocols etc.
Everything else is done in software at a higher level.
--
E-mail jkn...@ca.ibm.com
_________________________________________________________________________
The above opinions are my own and not those of ISM Corp., a subsidiary of
IBM Canada Ltd.
wclar...@my-deja.com
James Knott <jkn...@ca.ibm.com> wrote:
> Since when was any NIC sensitive to network class or even TCP/IP???
> Last time I checked, a NIC simply sent a packet from one MAC address
> to another, with no consideration for higher level protocols etc.
> Everything else is done in software at a higher level.
Oh god, it's been many many years since anything like this was on the
market. Some cards *are* aware of TCP/IP in some sense, by means of
their drivers. You are correct, AFAIK, in asserting that the NICs
themselves [we're talking hardware here] are pretty dumb. If the driver
refuses to put the card into the proper mode, and writing a custom
driver is out of the question [as it was most of the time about ten
years ago], then if the driver doesn't want to cooperate the card won't
either.
Sorry to be unclear about that, but most people I know don't distinguish
between the NIC and the drivers. I tend to just think in terms of
`ifconfig` affecting the card itself, but in reality it of course does
not.
bill davidsen
| Oh god, it's been many many years since anything like this was on the
| market. Some cards *are* aware of TCP/IP in some sense, by means of
| their drivers. You are correct, AFAIK, in asserting that the NICs
| themselves [we're talking hardware here] are pretty dumb. If the driver
| refuses to put the card into the proper mode, and writing a custom
| driver is out of the question [as it was most of the time about ten
| years ago], then if the driver doesn't want to cooperate the card won't
| either.
The last time I saw that I was an "ARPAnet technical administrator" if I
remember correctly. Not bloody recently, for sure.
--
bill davidsen <davi...@tmr.com> CTO, TMR Associates, Inc
The Internet is not the fountain of youth, but some days it feels like
the fountain of immaturity.