Comments on: Pageless in Podunk

Question: Pageless in Podunk

words1 — Mon, 20 Feb 2006 21:47:10 -0800

Particular web pages won't load for me -- nothing exotic, normal stuff like the Washington Post, certain pages at Amazon. The WaPo site just loads the logo bar at the top and then stops dead. Worst part is that it doesn't matter what browser or OS I'm using. What's going on?

We have three computers: XP, OS X 10.4 and Linux (Ubuntu), all hard-wired into a router fed by a wireless bridge connected to a parabolic dish pointed at a relay on the water tower in town two miles away. Believe it or not, this setup usually works fine.

But for the past week, I've been unable to load pages properly -- some pages (e.g., Digg) will load halfway, others (WaPo) barely even start before pooping out. Happens in Firefox 1.5, IE and Safari. Router firmware up to date, no problem fetching mail or even streaming music, but this is driving me nuts. I checked the router configuration for any sort of filtering (none), reset everything repeatedly, but no joy.

One possible clue is that last week at one point we could fetch mail just fine, but our web access was absolutely dead for about 12 hours. I haven't been able to get through to our ISP today, but they may not be much help when I do, so I'd appreciate any insight folks may have so that I can ask the proper questions when I do get through to them.

Any suggestions?

By: helios

helios — Mon, 20 Feb 2006 21:56:31 -0800

Do you know if your ISP uses a proxy server to cache web content? Next time you have web problems, can you test your browser using an anonymous proxy that's on a non-standard port (something other than 80, 8000 or 8080)? If the anonymous proxy works, I'd guess it's your ISP's fault.

If you still have problems, it could have something to do with multiple simultaneous connections made by your browser (that is choking the router or bridge). If you turn off images and javascript in firefox, can you browse the web without any hiccups? If so, you could adjust network settings in firefox to open fewer simultaneous connections.

By: words1

words1 — Mon, 20 Feb 2006 22:05:57 -0800

Turning off javascript and images doesn't help. I thought Fasterfox might be the problem, but I disabled it and it didn't help.

How would one use one of those proxies in, say, XP?

By: Good Brain

Good Brain — Mon, 20 Feb 2006 22:09:35 -0800

Can you adjust the MTU size for the wireless connection?

By: Good Brain

Good Brain — Mon, 20 Feb 2006 22:10:38 -0800

I should have said: If you can adjust the MTU size for the wireless connection, try reducing it by 16 bytes or so. You may also be able to set this on the client side, but I'm not sure how.

By: helios

helios — Mon, 20 Feb 2006 22:11:20 -0800

in firefox: tools -> options -> general tab. Click the "connection settings" button. click the "manual configuration" radio button. Under HTTP proxy, enter the IP address in the first field and the port in the second. so 123.234.234.123:2389 would be 123.234.234.123 in the first field, and 2389 in the "port" field.

By: puke & cry

puke & cry — Mon, 20 Feb 2006 22:13:43 -0800

in firefox, tools > options > connection settings > manual proxy configuration - enter http proxy and port.

By: RichardP

RichardP — Mon, 20 Feb 2006 22:55:10 -0800

I suspect the problem is PPoE or IP Tunnelling in combination with broken path-MTU discovery due to erroneous ICMP filtering (also here).

As an experiment, on the Mac OS X machine (and assuming the Mac is connected via Ethernet to your wireless router), try the following:

From the Apple menu, choose System Preferences.
Select the Network pane.
From the Show pop-up menu, choose Built-in Ethernet.
Click the Ethernet tab.
From the Configure pop-up menu, choose Manually (Advanced).
For Maximum Packet Size (MTU) select the radio button labeled Custom.
In the field to the right of Custom, enter a new value lower than 1500 - I'd try 1492 or perhaps 1484.

If this fixes the problem, you can either talk to your ISP to get the ICMP filtering fixed or set the MTU on all of your machines. For the Linux machine it would be something like "sudo ifconfig en0 mtu 1480", but off the top of my head I don't know where the MTU setting is adjusted in Windows XP.

By: theredpen

theredpen — Tue, 21 Feb 2006 05:29:33 -0800

Pretty sure no-one cares, but I'm on OS X and having a similar problem. Are you on Comcast cable? I figured there was network traffic, but our PC seems to be getting through. I wonder if they changed something. Which is what I wonder when I don't know what is causing something.

By: jellicle

jellicle — Tue, 21 Feb 2006 05:37:40 -0800

In my experience, "web pages loading halfway" without other problems almost always means that there's a broken ad server out there somewhere. Is it the case, for example, that all the pages you have problems with have Doubleclick ads on the page? They are notorious for such problems.

By: words1

words1 — Tue, 21 Feb 2006 10:14:30 -0800

By: RichardP

RichardP — Tue, 21 Feb 2006 14:03:43 -0800

So I spent some time talking to my ISP, and they're having the same problem with the WaPo site. Traceroute stops at gar1-p360.abnva.ip.att.net and nothing really loads. At least I know it's not my setup.

I don't think the fact that traceroute stops at gar1-p360.abnva.ip.att.net when tracing to www.washingtonpost.com is necessarily an indication of a problem in your ISPs network. For example, I can load www.washingtonpost.com fine, but a traceroute from my host to it also stop at gar1-p340.abnva.ip.att.net:

5 so-4-0.hsa2.tustin1.level3.net (65.59.168.1) 9.622 ms 8.847 ms 9.093 ms
6 so-1-1-0.mp2.tustin1.level3.net (4.68.114.21) 10.837 ms 9.089 ms 12.534 ms
7 so-4-0-0.bbr1.dallas1.level3.net (4.68.144.1) 42.399 ms 42.270 ms 41.293 ms
8 ae-23-52.car3.dallas1.level3.net (4.68.122.47) 43.859 ms 43.424 42.577 ms
9 192.205.32.221 (192.205.32.221) 40.548 ms 40.688 ms 40.341 ms
10 tbr2-p014001.dlstx.ip.att.net (12.123.17.86) 74.939 ms 72.976 ms 73.614 ms
11 tbr1-cl6.sl9mo.ip.att.net (12.122.10.89) 73.687 ms 74.041 ms 72.433 ms
12 tbr1-cl4.wswdc.ip.att.net (12.122.10.29) 72.558 ms 72.354 ms 71.862 ms
13 gar1-p340.abnva.ip.att.net (12.123.217.10) 89.204 ms 71.458 ms 70.859 ms
14 * * *

Since ping's to www.washingtonpost.com also fail for me, I suspect either ATT or Washington Post is performing overly aggressive ICMP filtering on the route to Washington Post's web farm, perhaps as an attempt to mitigate a denial-of-service attack. Did you try my MTU suggestion?