Why do they do it?

Probably due to so many clueless Windows users failing to do security updates and unknowingly running an easily infected web server or opening infected e-mail, that in turn spreads those infections. There are still infected machines trying to spread the old Code Red and nimda worms.

My ISP only blocks a few ports (related to MS Blaster worm, etc.), and not port 80. If I open my port 80 in my firewall, apache gets almost constantly hit by those infected Win boxes. So I resorted to a default virtual host that logs separately, and gets log rotated to the bit bucket. The only virtual hosts with any content are my no-ip.com names (dynamic DNS) that I mostly use privately. The worms can't find those.

There ain't that much bandwidth; around here, it's beginning to be pretty clear that Qwest has oversold their network.

Folks at my ISP have been collecting statistics on throughput vs. time of day and the customers are getting the speed they pay for only if they're online at 3AM.

I ran a test downloading a 1MB file from a server on my ISP's farm every half hour for a full week and found that I was seeing 20KB/s transfer rates at 6PM and 450KB transfer rates at 3AM (on a 5Mb/s line). Other customers running similar tests are seing similar results.

mike3 wrote in part:

Not all do, and even those that do often don't enforce this policy. Even though it would be trivial to do so.

There isn't as much upsteam bandwidth with ADSL. That's what a server would use.

Well, all bandwidth costs. But commercial links OC3+ are usually sold symmetric. Since home users actually receive ~10x more than they send, this gives unused send bandwidth that can be sold as hosting.

Yes, they usually charge biz users more.

Marginally. With a no-servers policy, they can shutdown most inbound ports when under attack.

You're talking OpenBSD. Yes, they're tough.

They can easily scan for port, or review their logs and find you.

-- Robert

So even if I had a super-secured server, would they still try and shut it down?

mike3 wrote in part:

Might well, if they don't want to support the bandwidth it takes. A lot of the ISP is run by marketing, who think in terms of products and strongly avoid commoditization.

-- Robert

It's not greed, it's just much more expensive to *support* a person who is running servers. So they want extra compensation for it.

You may be the most competent, responsible person in the world who would require no additional help or support no matter what you're doing, but unfortunately, they have to base the amount they charge you on the average cost to support a person who wants to do what you want to do.

DS

Because it would cost them money to enforce the policy. The purpose of the policy is so that if you call them with a problem, because you have no idea what you're doing, they can answer "we don't support that, if you want to do that, pay extra for a premium account".

It's largely a support cost issue.

DS

If it somehow came to their attention, they might feel they had to in because you're kind of thumbing your nose at them. My bet would be that, basically, they don't really care. However, you really shouldn't do business with a company that's not providing you the service option that you actually want. You'll be a lot happier when you're not doing something you agreed not to do.

DS

Two reasons... support and bandwidth. Both expensive. You *can* run servers if your ISP offers a commercial account - for more money, of course.

-Frank

Please see:

Exactly. Mainly bandwidth. Bandwidth is expensive and one or two people running servers can completely over run a network. There are some ISP's who permit servers, but will watch the traffic and if it gets to high, give you a warning.

These links you posted are all about "liberty and freedom of speech" (as quoted from the second link). The OP's question about prohibiting servers on consomer connections is not related in any way to liberty or freedom of speech. It is all about money. All. ALL!

-Frank

It has much more to do with the fact that they don't want you saturating their network with high-volume traffic applications. For high-speed access to make sense economically, there's no way they can give you the same "fat pipe" privileges that they do to businesses. If everybody did that, their fast network would slow to a crawl.

But most every provider has a business division, and some have even entered into joint ventures with companies like Lightpath that use their existing fiber optic network to give you really fast speeds. You just have to be willing to pay for it. And it's very, very expensive. But it's business class and guaranteed.

Yes and no, IMO. For example, if that were really the issue, they could simply place bandwidth limitations on the servers you could run. Also, they have an enormous amount of inbound traffic from the Internet and very little outbound traffic even though their internal and external links have the same bandwidth in both directions, so server traffic is extremely unlikely to cause them performance problems.

DS

They do limit your bandwidth. IE: 1.5mb down/256kb up. The real reason they prohibit servers is because they oversell their service. If too many people run servers and use all their provisioned bandwidth then the system would bog down. So in other words, you pay for service you only use maybe a few hours a day.

This brings up another interesting question: why does a dedicated-access line that you can run servers on like T1 (1.544 mpbs) cost 10x+ ($400 vs $40/month) the cost of a non-dedicated line of similar bandwidth (like DSL (it can sometimes get T1-level speed))? Is it because they have to run a line to your place instead of just hooking up to a preexisting network? network?

It is a combination of many factors, but the largest one is most likely the increased cost of the transport. If it was installation cost, you'd find the price going down as the length of the agreement goes up. (And sometimes it *is* installation cost and the cost *does* go down as the committment time goes up.)

DS

What exactly do you mean by the "transport"?

A T1 is a more expensive circuit for a telephone company to provide you than a DSL link. T1s were not designed to carry packets and as a result, your individual circuit has to be kept individual for longer than is necessary with DSL. (I'm assuming a classical HDLC T1 rather than, say, ATM on a T1, which helps to reduce transport cost.)

DS