Comcast works well if you buy your own modem. This allows you to actually pick one based on specifications/reviews and they will generally last long enough to save you money on rental fees.

With the exception of a brief interval on Knology, I've used Comcast internet service since they bought out the AT&T @Home service that I was on.

With Comcast, I haven't had any real issues with service reliability or customer support. At least not since I decided to skip the rental modem lottery and buy my own from Newegg. Prior to buying my own modem, I had a service tech out 2-3 times per month to tell me that there was nothing wrong with the installation that would cause me to lose my connection several times a day.

With Knology, however, I had nothing but terrible service and horrendous customer support. Comcast seems to be willing to invest the time and expense in maintaining and upgrading their networks, both the local systems and the national backbone. Knology in my area, however, has a very flaky local network ( to the point that customer service tried to convince me that 25% packet loss is normal and acceptable ) and that local network connects to the rest of the internet with an extremely oversubscribed link.

Some of them will dump if they're not careful with their pointers.

Must you ruin my optimism?

I just switched to Comcast to get away from Knology and take advantage of their new DOCSIS 3.0 rollout in my area. I'm currently getting around 35/9 (sustained over several gigabytes) on their 22/5 service. I know it's not going to last like that but for now I'm happy deluding myself.

Although I will say that Comcast will have to put in a serious effort to deliver worse service than Knology. With Knology I was paying for 6Mbit and getting 1.5 to 2.0 on a good day and enough packet loss to make bandwidth meaningless on a normal day. My leased modem had a manufacture date from last century. Service appointments were silently rescheduled to another day or canceled entirely. Issue tickets would randomly disappear from their system. And in order to pay my bill by phone I had to call them to give them my CC#, get the confirmation code, hang-up and call back with the confirmation code that I just received to inform them that I just called them 2 minutes ago to make a payment.

So as long as nobody from Comcast breaks into my house at night to violate my cats, I can't see them being any worse than the "competition" here in South Carolina.

If I had mod points, this would get a +Funny.

The problem with developers and statistics isn't that we don't appreciate how complex statistics can be, it's that we don't know what to do with the pretty charts and graphs statisticians produce for us. Our primary concerns usually have nothing to do with the standard deviation of transaction times. We're concerned, first, with the correctness of the response and, second, whether or not it fails gracefully. There are, of course, exceptions in real time applications.

Most performance graphs are not going to have a nice bell curve to them anyway. They're going to bias heavily at the minimum time to follow the shortest logic path and work their way up. There will be spikes and plateaus where different paths are taken through the code logic. There will be outliers where something failed in the code. There will be outliers where something random happened between point A and point B ( a packet collision on the way to the database, a bad cache miss that will correct itself, garbage collection triggering, etc.). And there will be situations like what TFA's author is actually looking for, load based failures. Standard deviation is useless if you're not working with samples that have a normal distribution.

Outliers like that indicate to me that something broke for that sample. If possible, I'll investigate whatever request caused that outlier so that I know whether I'm facing a code/logic bug, a known low performance request, a load based failure or just some random influence outside of my control. Just to clarify, a known low performance request would be something like a database query calling a stored procedure to generate a large report.

Present day astronauts recycle their urine for drinking water. The space and plumbing requirements are less than if they were to stock the ISS with plenty of water to last between resupply runs. Star Trek, having the tech to break apart and rearrange matter on a molecular level, recycles all waste, solid and liquid, for materials to make fresh provisions. This doesn't mean that they crap on the kitchen counter, it just means that whatever you flush goes to a system to recycle it.

Maybe if you were trying to tie them together on the same mainboard. But over TCP/IP not an issue.

If you do not have the hardware to support Windows 7 then don't upgrade. If your graphics card is the only thing holding you back then take a stroll over to Newegg and start upgrading.

Complaining about hardware that was designed for Windows XP not working in Vista/Win7 is really akin to complaining about hardware that worked fine in Win95/98/ME not working or working well in XP. Eventually you have to upgrade hardware to run modern software. If you think ATI is choosing to end support for a legacy product too early, feel free to buy a nVidia card to replace it. Eventually though, nVidia will discontinue support for their DX9 products the same way they no longer support DX8 products in Vista/Win7. Some point after that, they'll discontinue support for the GT 285 for some future OS.