I made up my mind on my prior shared hosting vs. roll my own dilemma.
I’m going to use shared hosting for everything except two things:
Even better, we’re going to use Windows Server 2008. Ubuntu and Linux are such pains to manage.
I am at a crossroads with my neighborhood association’s electronic presence.
Currently it consists of:
I have an opportunity to use my own server to host all these features and some more. At first glance, this seems appealing, but it is so cut-and-dried:
| Feature | Shared Hosting or Google Groups | Ubuntu server |
| What I have to maintain | Web applications at 1and1.com and the five Google Groups. | Web applications plus:
It all runs as a Hyper-V virtual machine on a Windows 2008 server that I don’t have to maintain. |
| Technical prowess of maintainer | Generally expert maintenance. | I am well experienced and degreed as IT staff and can work my way around Linux systems pretty well, but I am not a Linux expert. I will have to trust Ubuntu’s and its supported packages’ default configurations and heavily rely on the aptitude package manager.I only have academic knowledge of advanced security methods like chroot. |
| Support quality | 1and1 can be flaky, but with enough piddling with support, the problem will be solved. Google has no support to speak of, but their stuff usually just works. | |
| Security | Tight security commensurate with shared hosting. | |
| Error handling | Difficult since I cannot see the Apache logs. | Since I have full control over the server, I can control my own Apache logs. |
| Flexibility | Limited to what 1and1 or Google Groups will allow me to do. That being said, I still have tremendous flexibility within their packages. | Full control, but in some cases, like with Mailman, the best available software is primitive and feature-poor (see below). |
| Control over data | Web hosts occasionally terminate hosting for arbitrary reasons, although likelihood is probably minimal given the uncontroversial content. Furthermore, there are avenues of recourse since I am a paying customer.Google Groups is a wild card. We are not paying, and their standards for what constitutes a spammer is arbitrary. If too many complaints gets lodged against any of my 5 groups, they could be deleted. I guess as long as I regularly download copies of their respective membership lists, I have an “out”? | Full ownership (if backups regularly happen). |
| Hardware redundancy | Reasonable. I have only encountered a few brief outages with 1and1. | Zero, although server owner plans to sell some services he will provide on the Windows 2008 master machine, so he has incentive to keep it working. |
| Data storage limit | 10GB currently, but a 120GB plan is just $1 more per month. | Currently 34GB. Additional space may be trickier to add and could more closely parallel actual cost of equipment (versus $1 per month). |
| Outages | Both 1and1 and Google Groups have had minimal outages. Not sure if either would get past “three nines,” but that is plenty for my purposes. | Outages and duration thereof could be more likely given lack of redundancy and overbooked and only mildly competent support staff (me). |
| Passing the torch (I won’t be system admin forever!) | Easier since I can turn over Google Groups management to someone else and could theoretically send all stuff hosted at 1and1 to someone else for hosting at his own shared account. | More difficult because, again, I will be the only support staff. I can act in a server admin role, though. |
| Server speed | 1and1’s speeds usually acceptable. They used to drag a lot but they are getting progressively better. | Very fast. |
| Ability to support upcoming camera monitoring project | Will be a challenge with 1and1 due to how there is only 1 FTP username/password per customer account. | Fully customizable. |
| General email list functionality | Pretty good with Google Groups for announce-only or run of the mill lists. Some users have complained of not receiving emails, but nobody has been able to substantiate problems, so these may not be valid gripes. | Mailman appears to be the best option, and it’s simultaneously primitive and too option-rich.For example, it’s not possible to set up an announce-only list. Yes, you can change some configuration options to emulate one, but you still have to set each new user’s moderate flag.Mailman’s user interface sucks! Give me Lyris any day!
Majordomo is not an option. I will not run a program that hasn’t been developed in 8 years. Phplist may be an option, but that is yet another program to support! And it’s not supported directly by the Ubuntu project. |
| Emergency SMS alert list functionality | Google Groups probably isn’t right for this. There’s no way to turn off the footer, so even a blank message will exceed the 160 byte SMS message boundary. | Mailman may be an option (albeit with the above difficulties concerning announce-only lists), but phplist may also work. But that would be yet another stinking program to install! |
| Backup | Regular. | Probably sporadic and not automated. |
| Time commitment | Baseline. | Baseline plus time to maintain all components and support system. Could be significant if package upgrades increase commitment. However, if packages are “set and forget”, then commitment can be trivial. |
The argument gets even more abstract.
What’s the best way to ensure I can pass the torch to someone else? To not even use my own software. That means the correct option is “none of the above.”
There is already a service called neighborhoodlink.com. Some adjacent neighborhoods use this company. For example, Little Forest Hills. Why not use this? Sure, it’s kind of ugly. Sure, it’s feature poor, and it looks kludgy. But will it “get the information out”, do it for the least possible effort, and increase continuity? Yes, yes, and yes!
This is a tough decision.
A huge error is in many articles discussing ICANN nascent TLD rules. Here’s a quote from PC Magazine, which should know better:
The additional domains will also probably accelerate the shift to IPv6, an expanded IP addressing scheme that will provide roughly 3.4×10E38 IP addresses, or ten billion billion billion times more than those provided by IPv4, the current scheme. (source)
In fact, there is no direct relationship between IP addresses–abstract numbers–and domain names–the human-friendly, text-based names.
IP is the addressing system of the internet. Every internet-enabled device talks from its own unique IP address to the unique IP address of another machine. It’s just like when you send a postal letter, you sent it “from” your house’s own unique address, the return address on the letter, to the unique address of the recipient.
When you type a web site name in a web browser, such as www.smu.edu, the browser looks up the web site’s IP address. The browser then “talks” to that IP address.
It’s similar to correlating a person to his cell phone number. If I want to call John Smith, I can’t dial “John Smith” in my phone. I have to look up and dial his phone number instead. During the call, I know I’m talking to John Smith, but the phone is simply communicating with an abstract phone number.
IPv4 is the current IP addressing scheme. The is, under the most dire predictions, all available IPv4 addresses will be used up in a few years. In that event, no new devices can use the internet.
An analogy: Suppose a road is very long, and road’s houses have three digit addresses: 001 to 999. With that scheme, only 999 houses can be on the road. If the address changes to 6 digits, the road could allow 999,999 houses because addresses range from 000,001 to 999,999.
IPv6 addresses are like adding those additional digits. In fact, it has so many digits that each person could have fifty octillion (50,000,000,000,000,000,000,000,000,000) IP addresses before that system becomes exhausted.
(Truth be told, the predictions of IPv4’s collapse are grossly exaggerated. Simple workarounds are already available that could allow IPv4 to work fine for a long time. And because of the way it assigns IP addresses, IPv6 in fact cannot deliver nearly the number of addresses advertised; as is the case with IPv4, but for different reasons, there will be significant numbers of unusable addresses. But it is true that IPv6 really does have several orders of magnitude more addresses than IPv4, and IPv6 also has several convincing technological advantages that justify its use.)
Back to the point of this article: the IP systems’s current address space crunch is a technical artifact of the IP system. It has no relationship whatsoever to the domain name system. Domain names are merely pointers to certain IP addresses. Nothing more, nothing less.
Today the ICANN voted to relax rules governing top-level domains (TLDs). TLDs are the last part of a server address and commonly end with .com, .net, .org, .biz, .edu, .gov, et al.
Currently, ICANN arbitrarily and capriciously regulates creation of new TLDs. For a great example, the .xxx TLD was debated, authorized, then arbitrarily rejected.
ICANN’s newest decision loosens things up. An institution wanting its own TLD will “merely” need to pony up several tens of thousands of dollars and prove competence to manage its own TLD. For example, if I had enough spare change and a good IT organization, I might buy the .cambre TLD and start selling .cambre domain names. (Think of the prestige of owning www.aren.cambre. Totally awesome!)
This will have several effects:
I am excited by the ICANN’s decision. Finally, the ICANN injected sense into the domain name system.
I’m writing this post on the Mac. I’ve come to like and dislike things about the Mac.
To sum it up, I do not understand the fascination with Apple. It seems to be driven by a misguided response to Microsoft Windows Vista. I really feel Vista is overall a superior OS.
Mac likes: