Bus error: Jake Billo’s weblog

While it may seem like all I write about these days is Rogers, it’s really the only thing I’ve been dealing with on the service provider front. All my other corporate relations have been going well: I pay people money and they provide a service without bothering me unduly. (I must congratulate the wireless business for their 6GB data plan extension and forthcoming reasonably priced data packages, although one could make the case that Telus and Bell really forced them into it.) This time, it’s about the Internet side of the equation.

Beginning July 18th, Rogers began implementing a provider-wide SiteFinder-style service, where users are redirected to a “search” page with sponsored results for mistyped and nonexistent domains. On a technical level, I fundamentally disagree with this change: it breaks the concept of NXDOMAIN (a useful “domain does not exist” response) and makes things much more difficult to troubleshoot with respect to network architecture. The only reason I haven’t bitched and whined about this much earlier is that I’ve been using OpenDNS for completely unrelated reasons. It was only when my roommate Alex complained about VPN connectivity that I actually looked into the issue.

It turns out that Rogers’ marketing effort completely bricks internal domain resolution for a lot of common VPN clients, including the default Windows XP offering. So if your company, like many others, has internal domains such as corpweb.example.com, Rogers’ search will open up with the terms “corpweb example” at the minimum. This practice has data exposure implications: not only does Rogers now know about an internal domain you’re trying to access, but a third party provider like Yahoo now knows.

If you were an employee of a competing search engine and trying to VPN from home, Yahoo would now know something about your internal network structure; this is bad news all around. Hitting a favourite or quick launch link to corpweb.example.com/livelink/llsapi.exe?doc=Network_Security_Breach_Sept0408.doc would reveal the choice of LiveLink as a corporate CMS, a dependence on Microsoft Word and a document detailing a potentially classified incident.

OpenDNS isn’t any better by default, either. They redirect search results and mistyped domains, and in the process intercept VPN traffic. To get around this, you have to create an account and blacklist corporate VPN connections from “helpful results” on a per-domain basis. The solution also involves downloading and maintaining a dynamic IP address update client, or setting a Tomato-enabled router to perform the same task.

What I’ve done for now is listened to the accurate advice on trevoro.ca and changed my primary Rogers DNS server to an unadvertised IP address: altdns.rnc.net.cable.rogers.com, or 64.71.255.202. This server seems reasonably quick for name resolution and returns proper responses when a domain is not found, allowing VPN software to resolve internal addresses.

From a HowardForums post, I’ve found out that the Canadian Internet Policy and Public Interest Clinic is seeking comments from people affected by Bell or Rogers’ throttling. This refers to Bell’s global limitation of about 30K/s for BitTorrent, and Rogers areas affected by Ellacoya/Sandvine boxes.

I’ve written Robert Hester an email about my own experiences on TekSavvy, which I’ll repost below. If you’ve been affected by this nonsense, it might be worth your while to write in before a decision is made in September.

Hi Robert,

I read a post from a CIPPIC alias on HowardForums (http://howardforums.com/showthread.php?t=1406593) and figured I’d contribute my experiences with Bell’s internet throttling.

I’m a Computer Science student in Waterloo, Ontario, and currently subscribe to TekSavvy - a Bell DSL reseller. I chose them because I consider Bell Sympatico service to be substandard in the market: Bell’s technical support is lacking at best, they limit bandwidth usage to 60GB per month, and they don’t offer additional features such as static IP addresses.

Bell has applied throttling to their wholesale customers including TekSavvy, specifically for P2P protocols such as BitTorrent. When this network management is active - which I find tends to be more than twelve hours out of every day - my BitTorrent download speeds are limited to 30-50KB/s. This is a far cry from the 5Mbit (~500KB/s) service that I pay for.

What really disappoints me is that an independent reseller has to suffer because Bell unilaterally decides to squash competition. Unfiltered P2P and unlimited bandwidth usage are key competitive features that I’m willing to pay for. When a third party like Bell decides that they can’t compete and uses technical restrictions to bring other services down to the same level of reduced functionality, I’d go as far to say it’s an unfair trade practice. In my opinion: if this practice is not currently illegal, it should be.

I have no objection to reasonable network management, but reducing download speeds to 10% of their potential based on a transmission protocol is far from reasonable.

Thanks for your time and consideration on this important issue. Please feel free to contact me at the email address or phone number below if you require anything else.

Jake Billo
jake@jakebillo.com
http://jakebillo.com/
519-279-4009

As the informal network weasel in my new place, I get the wonderful joy of troubleshooting malfunctioning appliances and making sure that the router eats as few Xbox Live sessions as possible. Since I’m just lazy enough not to want to set up a Linux routing box, the current approach for networking is two connections into two routers:

• Rogers Hi-Speed Internet Extreme (95GB cap), into a Linksys WRT54GL running Tomato 1.19 firmware and
• TekSavvy, unlimited cap, DSL dry loop, into yet another Linksys WRT54GL running Tomato 1.19

The main server with two network cards accesses the Internet over the TekSavvy line, using a combination of manual interface metric settings and a MAC address block at the Rogers router.

It’s not the TekSavvy line that’s been giving problems, though - and the Rogers connection is solid, even with four computer science types all wanting their pornography and HD movies updated Java Development Kits seven times a week. It’s the Rogers DNS servers that cause problems looking up domains - I’ll often receive 60 to 120 second timeouts just seeking a match for facebook.com. Boo, you fail!

The solution is to switch DNS services to OpenDNS at the router level. Tomato provides an excellent internal DNS cache service, which still allows Linux systems to access internal hostnames - and OpenDNS returns lookups reliably and without fail. The price you pay for this is a page of sponsored search results on a domain typo or non-existing hostname, but this is fairly similar to how most browsers function anyway.

To activate OpenDNS in Tomato firmware, you can change the “Static DNS” settings in your router administration panel. On default configurations, the address is 192.168.1.1 with username root, password admin. Then it’s just a matter of adding the server entries 208.67.222.222 and 208.67.220.220:

There you have it - DNS that still resolves local systems, but is significantly more reliable than the ISP-provided service.

It’s happened, and people should start getting upset. Rogers has begun injecting packets to indicate that people have reached their “usage allowance” into arbitrary websites - which curiously does not include Google, likely due to the previous public outcry when test images appeared on Ars Technica.

The top frame on nearly every website obscures text and causes web pages to load extremely slowly. In my testing, the content loaded from 64.71.251.10 and often required websites to be reloaded before they would display properly. Unlike preview screen captures of this technology, there is no way to select “don’t display this message again”.

For interested individuals, the “acknowledge” address is http://64.71.251.10/isnsack.pl, which uses a JavaScript function to submit a CGI request.

Rogers users last week were delivered letters indicating the provider’s intent to start implementing usage caps of 60GB and 95GB for the two standard tiers, Express and Extreme. Their misinformation site contains some interesting usage calculations, but essentially muddies the waters by describing activities in an “or” context: for example, you could download 24 high-definition videos (at about 4GB each) OR download 24,300 songs.

I’m admittedly surprised at their references to BitTorrent and high-definition movie downloads, but frankly that’s what their target audience is with this particular site: heavy users that they hope to coerce into keeping under a specific 95GB point.

The problem for consumers with this approach is that it’s no longer just the top 10% of users being targetted - a 95GB cap means that the top 10% of users will be using close to that amount, and it would be trivial for the ISP to reduce the figure in the future, even as more bandwidth intensive applications emerge.

Good news, though: extra bandwidth usage charges top out at $25 extra per month, which is a fee I’ll likely be paying part of every month upon my next move. (I also have full intent to purchase a TekSavvy unlimited account as well.)

From a network management perspective, I understand the need to keep subscribers and abusers of the service under control. There have always been provisions in every Internet contract for disconnecting people based on overuse, even back when dialup connections were standard. I’m irked that something previously touted as “unlimited” has gotten to the point where a hard cap and extra charges are necessary. Having heavy users on your network comes with the territory of having users like Granny checking her email once a week.

Rogers injects a usage frame into a webpage.

Admittedly, my Internet usage patterns tend to encompass newsgroup downloading and hefty SCP/SFTP transfers - on a regular basis working for IBM, I’d have to prop the latest Eclipse builds from Ottawa down the home line. Combined with the rest of the mandatory software suite, the nightly package could be 5GB.

One thing that will be interesting to see is Rogers’ deep packet inspection and rewriting - the usage message that caused issue back in December. They have an example image shown on one of their magazine property websites.

As soon as they start to insert HTML arbitrarily into webpages, that’s where they’ve crossed the line.

Are you going to change your internet usage patterns as a result of these developments? How much bandwidth are you using on a monthly basis?

From BroadbandReports, crossposted to ev98:

New Rogers speed increases are finally here after new price increases. For those curious, I’m on the Extreme profile and am currently pulling 900-920K/s speeds (8MBit) downloading. There’s about 2K/s increase in my upload speed, but I’ve yet to perform any serious testing. Express users (standard high speed) are reporting about 7MBit down with a slight increase in upload speed.

Rogers’ site doesn’t have any updated information, as expected, but this is good news for now. The downside is that they’re probably going to start enforcing their caps of 60GB/100GB per month, which is unpleasant news for heavy users. We’ll see what happens in the next little while.

To activate the changes, unplug your cable modem for five seconds, then plug it back in. This lets the modem download the new speed profile. My IP address hasn’t changed from before the speed upgrade, for what it’s worth.