ThisDomainWasFree Weblog

Update: How novel, I’m only chatting a little shit! You should read this. You can read my post if you like, but those two are a lot better.

I have been with Dreamhost nearly a year now and I feel that this is a reasonable amount of time to review them. A lot of people are suspicious about Dreamhost, and they have every right to be. Why? Referrals.

It cost me about $20 to sign up for the “Crazy Domain Insane!” plan, as opposed to the $119 it should have cost me. This is because I used a “referral coupon”. If someone ever gives you a Dreamhost referral coupon that takes off less than $97 then they are profiting from that coupon. You see, once you’re a member of Dreamhost you can create your own coupons redeeming up to $97. If you make a $50 coupon then you get ($97-50)=$47. So never, ever believe what anyone with a < $97 coupon says, they just want you to give you their money. This coupon business is what makes me very suspicious about Dreamhost, the vast majority of their sign ups result in them getting only $23.

Twenty three dollars does NOT pay for 20 GB of disk space and 12 TB of Bandwidth (1 TB per Month).
Or does it? The truth is that Dreamhost are so huge that everything comes to them at crazy wholesale rates. They’ve never publicly said how much it costs, but many times they’ve commented on how it costs them cents for every plan and despite the referrals they make profit.

Somehow it still seems too good to be true, right? Yes, that’s because one important factor is missing, this is processing. One Dreamhost server can host thousands of websites. On one server alone I see 1,139 user accounts, users may host as many websites as they like on their account so this could equate to tens or hundreds of thousands of websites. And handling them all is Dreamhost’s main problem. This is where they are known to get a bit tricky. They don’t tell you what their limits are, don’t tell you when you’re near your limit and are generally vague about the whole thing. Of course, if you do go over it they will kill your site and send you a nice e-mail suggesting you stop eating up their CPU and buy a dedicated server from them. Rumour has it that they do not impose a processing limit on sites that aren’t at the end of their 97 day “money back” period.

In all honesty, this is a problem for many hosts. Back in the day they served up HTML files and now they have to support PHP, MySQL, Ruby on the Rails, etc. etc. all of which can use up serious processing power. Controlling and limiting this is difficult.

Processing problems appear to be the main cause for a lot of anti-Dreamhost sentiment I’ve seen lately but people also take issue with the support. Frankly, up until recently I would have to agree with them. One support ticket took a week to get a response others misunderstood me completely and did not help. However, lately the support has really gotten its act together, it took 4 hours to get a reply from an enquiry about a broken MySQL server (which they politely pointed out was my fault).

So to conclude:

• Don’t believe anyone with a Dreamhost coupon for less than $97. They are lying about Dreamhost.
• Wholesale prices mean Dreamhost’s deals might be legitimate, but there is probably a little overselling.
• Processing is where the catch is, if you have a PHP site likely to generate 1TB of traffic a month .. get a dedicated server.
• Support and stability are getting better
• Every host has these problems

If you came here in search of a coupon then I have made one for you: FF9797OFF

This coupon gets you everything off that it can, to proof this I have attched a screenshot :

I’ve often wondered about portscanning the entire Internet.

There are 4,294,967,296 possible I.P. addresses, I know that a lot of these are unused or reserved for things like private networks (around 19,000,000 according to Wikipedia) but I think were I to scan the Internet I would want to try them all. Just for good measure.

After much deliberation I decided I wouldn’t be able to scan all 65,535 ports. I believe I would stick to port numbers 21, 22, 23 and 80. Just the four, to get an idea of what’s going on. It would be tempting to throw in 135, 139 and 445 just to see how many people still have publicly accessable Windows shares but that would almost double the amount of ports I would scan. I think that four would be perfect.

So after a bit of fiddling with nmap I settled on this, simple command:

nmap -P0 -p 21,22,23,80

I think it might require a bit of tweaking but that would probably just about do it. This would at least give me an idea of just how long this would take me. I tried google.com as my first target, to get an idea of how long responsive servers would take to scan

lag:~# time nmap -P0 -p 21,22,23,80 google.com

Starting nmap 3.81 ( http://www.insecure.org/nmap/ ) at 2006-04-11 17:25 BST
Interesting ports on 72.14.207.99:
PORT STATE SERVICE
21/tcp filtered ftp
22/tcp filtered ssh
23/tcp filtered telnet
80/tcp open http

Nmap finished: 1 IP address (1 host up) scanned in 11.724 seconds

real 0m11.731s
user 0m0.010s
sys 0m0.000s

11 Seconds. Oh dear.

I then tried an I.P. that I knew wouldn’t exist

Starting nmap 3.81 ( http://www.insecure.org/nmap/ ) at 2006-04-11 17:27 BST
Interesting ports on 123.4.5.6:
PORT STATE SERVICE
21/tcp filtered ftp
22/tcp filtered ssh
23/tcp filtered telnet
80/tcp filtered http

Nmap finished: 1 IP address (1 host up) scanned in 3.417 seconds

real 0m3.424s
user 0m0.010s
sys 0m0.000s

3 seconds, not too bad.

Now let’s assume we have the 19,000,000 non-existant/network I.Ps

19,000,000 * 3

Added to oh, let’s say that half of the I.P. addresses that aren’t reserved for private networks are active

1/2(4,294,967,296-19,000,00) * 11

And the remaining

1/2(4,294,967,296-19,000,00) * 3

It’s some seriously rought estimating, but it could give us a decent ball park figure.
O.K., I make that: 30,108,471,072 seconds

Which apparently works out a a whopping 3767~ years.

And by then IPv6 will almost certainly be in place. Crap.

This is for me, really. To install Perl modules without priveleges privileges do the following;

wget http://location-of-perl-module.com/perl-module.tar.gz
tar xfzv perl-module.tar.gz
cd perl-module
perl Makefile.PL PREFIX=`pwd`
make
make test
make install

Then, add to the top of the perl script

use lib “`pwd`/lib/”;

Where `pwd` is the directory you have write access to, i’m just being lazy.

I started a damn blog, and it’s even got “blog” in the name.
Bugger.
Oh well, as long as no-one knows about it..