The Answer Gang
ISP and Internet.From Neemc
Answered By Mike Orr, Jim Dennis
Hi all, I am Mike,
As far as I read your answers about ISP and Internet servers, I never found an answer about the difference between an Internet server and an ISP. I mean, what is the difference between an ISP and if I run my own Internet server which allow dial-in?
[Mike] Traditionally, "Internet Service Provider" (ISP) has meant an organization that provides an upstream access point to the Internet (via dialup, DSL, T-1, etc). This can be a professional ISP company or (in large companies) an "internal ISP" run by their own IT staff. And of course, the big backbones are "ISPs to the ISPs". One can informally call personal server offering dialup connections to a few certain people some kind of ISP.
AOL and MSN used to be private networks with their own content, along with CompuServe, Prodigy, etc. They offered something very different from "Internet access", which meant a connection to the Internet and then you were on your own to find content. But now AOL has combined the two, stretching the definition of what an ISP is.
At the same time, people used to get their webhosting and e-mail via their own ISP, but now they are increasingly using organizations who do not offer dialup access directly. And now there are new services like remote backup or "word processing via the web" (which, parenthetically, are a serious application of the old "multiplayer game" concept). Are these organizations ISPs? It's debatable. If you take this definition to its limit, anybody with a server offering any Internet service is an ISP. I don't think most people would want to go that far.
Of course every ISP has to connect with or through other ISPs. The "Internet" is an "internetwork" of connected networks. (I know that's a tautology, but bear with me here). The Internet is an internetwork of networks which all speak TCP/IP and have agreed on a set of common address delegations and naming services.
Thus their can only be one delegation of the address 198.41.0.6 on the Internet. Likewise the name www.internic.net must be delegated to one authority (which can publish various addresses, mail exchange and other types of records about that name).
Of course it gets more complicated than that in many cases. For example it used to be common to have gateways between the Internet and various other online services and networks. Once upon a time AOL, CompuServe, Delphi, Fidonet and various other networks were NOT "on the Internet" --- though e-mail and some other services (file transfers) could be accessed through gateways. The distinction between a "gateway" and an "Internet connection" basically boils down to the matter of addressing and naming. Those networks didn't have globally direct routable IP addresses (DRIPs) and (thus) their internal nodes couldn't have name/records in the DNS (domain naming services) system.
Of course many corporations use their own internal IP addressing (usually following RFC1918 which reserves 10.*.*.*, 172.{16..31}.*.*/12 and 192.168.*.* from use in the global address space). Systems on these networks use application level proxies or NAT (Network Address Translation: specifically IP masquerading) to access the Internet.
Proxies are also known as "gateways." I friend of mine likes to say that "if your connecting to the 'net via NAT then your 'NAT' really on the 'net!" We can think of NAT as a network layer "transparent" proxying service. (And there are some protocols can don't work through NAT, or requiring kludges to "mostly" work through it).
Of course some companies may offer you "Internet service" through a set of NAT gateways and/or proxies. I think that some of the cable modem services, wireless and even some of the DSL services are NAT'd these days. Whether they are "really" ISPs is a moot point. If you can get the specific services you need through these companies then they can be "service providers" for you.
Of course, I prefer a traditional ISP (one that gives me a static block of DRIP addresses, a route to my dedicated line, a dial-up line for fail over and remote use, and control of my own DNS and even control of my reverse DNS --- which is only a subnet of a class C in my case). I pay a much higher rate for these services than most DSL, and cable modem customers though I get a relatively paltry bandwidth for it --- only 1.44Kbps over IDSL. I don't complain because this gives me the flexibility to do anything "on the Internet" as any ISP or corporation (only slower than the big guys). Besides, I pay less than a quarter of what a few friends of mine used to pay for 56K frame relay leased lines only five years ago.
So, I would conclude with the following definitions (and let the hair splitters pull out their loops and razors):
An "internet server" is any host which is "on the Internet" and providing services to clients over it. Usually this means "public services" (such as most web, FTP, DNS, etc that you've ever heard of). I would even say that a "server" could actually be a small cluster of systems which are providing a service as a group (usually behind some sort of load-balancing NAT router) constitutes an "internet server."
An ISP is a company or organization which provides internet services (connection to the world wide Internet address space, routing and naming system) to customers (as opposed to their own staff, departments or satellite offices).
(Note that I *don't* count companies, such which provide a stack of dedicated remote access lines their employees and let them access the Internet through their corporate connection as "ISPs" in this sense.)
A "tier 1" Internet connection is one which has an AS (autonomous system) number, which is a sort of routing address that is completely different then IP addresses, and uses that for "peering" (exchanging routing data) with multiple other tier 1 internetworks.
A "tier 1" ISP would logically be a company which has a tier 1 Internet connection and which provides Internet connections to customers. (Note that some organizations which have first tier Internet connections are not ISPs --- some large educational, governmental, and military organizations, for example).
On the next tier down are ISPs and organizations that lack AS numbers and which don't peer; but which do have CIDR blocks of DRIP addresses. (CIDR is "classless interdomain routing" --- a way of aggregating blocks of numerically adjacent addresses so that they entire block can be managed with a single entry in the global routing tables. Such organizations might be "multi-homed" (given them some load-balancing, and failover redundancy through their own policy based routing).
The lowest tier of ISPs aren't multi-homed (don't have multiple independent Internet connections) and/or don't have their own address block delegations. They have to get their addresses from their ISPs and they have no redundancy (or they are completely dependent on their single ISP, even if they do have multiple dedicated lines to them).
Of course there are places where customers rent rackspace and connect their computers to an Internet connection, and power. These are colocation facilities. In some cases you're renting whole cages. In most cases they are providing UPS (uninteruptible power supply and generator power backups), physical security, 24 hour monitoring, and other services (such as access to network tape drives for your backups and terminal servers for remote console access to your systems). Colocation facilities allow you to bring your computer to the Internet (rather than bringing the Internet all the way to your computer). They are an imminently logical approach when you need high speed and low latency at moderate cost for your internet servers.
![[ Answer Guy Current Index ]](../../gx/dennis/answertoc.jpg) |
1
2
3
4
6
7
9 |
![[ Index of Past Answers ]](../../gx/dennis/answerpast.jpg)
 |
| 11 12 15 16 18 | ||
| 20 22 24 25 26 28 29 |