The pickings were a little slim for wired routers at the local shop. They've got a couple really cheap offerings, like $20, but they only offered 10MB ports. They've got a few mid-price offerings, like $50, which is what I picked up. And they have a few higher end options, over $100, which is what I wanted. I had looked on-line earlier, but I couldn't find the one I'd thought to look for. I settled on a Netgear  RP614v4, figuring that Netgear's fair gear; I've got a couple of their GB switches and they work just fine, and simple Internet NAT routing should be old-hat by now.

Apparently they can mess it up a lot.

I'm not sure it was the slowest interface I've worked with. I mean, I've been doing this a long time, so there must have been some old, slow web-based router interface before, when machines were still rated in megahertz clock speeds. Not that a router should be judged by the speed of the web interface it has, but really, it shouldn't be that noticable.

Also, the interface truly sucked. There's not a lot to configuring a SOHO router. There are a couple of interfaces to configure, and some options for DHCP. If the router offers any kind of filtering or forwarding, some interfaces for that, too. The layout of the information was pretty fair, but it's all clumsily wrapped in frames with awkward colors. And did I mention how slow it was?

Don't get me wrong, the router seems to work. I watched a slew of movies on Netflix using it over the weekend. And I've been successfully web-surfing an otherwise using the Internet through it. But while it's performance (outside of the web interface) wasn't a problem, its reliability was.

The real problem I had, and the reason those two things even came to my attention, was that the router seemed to time-out and "disconnect" from the Internet after some brief periods. Activity didn't seem to matter; in fact it stopped responding while I was actively researching how to find where to correct this by using Google, because their support pages (linked above) are also attrocious.

It seemed to be the consensus in the bits I found, while searching for connection failures that weren't related to DHCP or msiconfigured IPs, that there is some PPoE-related time-out feature that might be working even though I'm not using PPoE, nor that I was at all inactive. Maybe the time-out isn't related to activity, but some kind of session limit, for whatever reason.

The only bit related to time-out was with port-triggering. Port-triggering is used to dynamically open an incoming (from the Internet to the LAN) connection based on an outgoing (from the LAN to the Internet) request. Most often this feature is used by games or services where the outbound request will alert the router that the inbound service will be available; the software hitting the server will then allow the server to make a request back to the software (essentially turning it into a server...). I'm not doing anything that needed that, so no configurations were created. I did, however, change the time-out to "9999," as the on-page instructions suggested. This didn't correct my problem.

I even tried adding minutes-long schedules of blocking and specifying something in all of the spots that seemed like it might be trying to help the router justify having a connection. I had already configured a virtual host to do SSH my desktop; that always seemed to work, even when the out-bound Internet "timed-out," indicating that there wasn't anything wrong with the Internet connection, just the router's desire to route out-bound traffic.

How I tested was to notice when the routing seemed to stop, through normal use. I'd then re-try a failed page, failing again. I'd then turn to the machine that's directly connected to the Internet (firewalled using iptables, rest-assured) and SSH through the router back to the other desktop. After that (slowly but successfully) connected, I'd turn back to the desktop that used the router and again would try the page, and it would fail. I'd hit the router's admin page and "test" the connection, and all would return to working order for a few more minutes (not 9999, though).

Short of adding whatever software is on the disk that came with the router (none should be required, especially if the router has a web interface for configuration) and seeing if that corrected the problem (which would probably require having that software always running, to tell the router I wanted to use the Internet; and I'd have to put it on a Windows VM anyway...), there isn't anything else I can see that I could configure to tell the router that it should never stop handling out-bound requests.

I pulled the router and stacked it atop its packaging; I'll gather and box it properly later before I take it back (I hate returning things, but this is beyond my "shoulda known better" lesson-money range) in the next couple of days.

I tried plugging the old router back into place. It wouldn't finish booting. It must have either been totally insulted by getting replaced, or perhaps the equipment gave out after cooling down. It's the first time the thing's been off for more than a couple hours since it was pressed into service. I was going to put up with its throttled Internet connection for a while until I could replace the Netgear, but I couldn't not have anything.

The only other router I have in the house is the wireless. It isn't in service as a router, though, just a WiFi access point. Plus, it's centrally-located on the main floor of the house, giving the second floor a fighting chance at getting a strong signal. I pulled it from its station behind the TV, popping an Ethernet extension dongle in its place so the PS3 would still have access, and took it downstairs. I plugged it into its new place, set it where the antennae wouldn't be physically interfered with, and returned to the office.

It was already booted and waiting. I turned on the DHCP server, giving it the same range of addresses, hoping nobody would conflict, and set the gateway and DNS to its LAN address, instead of the other's address. I tickled the Mac's AirPort, and it got a new address from the new server, with the right other settings. I configured its WAN interface with a static IP, the one of the first router (off for a few days now), but it didn't work. I tried in vain the one used for the Netgear, which didn't work either.

I've run into this before; although I have a block of static addresses, my ISP "protects" me (or them, I'm not sure) by locking the IP to a MAC address for a while. Of course, the MAC can be "spoofed" (which is a bad-sounding way of saying "set manually"), but I didn't have the other routers handy to find or otherwise divine their MAC addresses. On network cards there's usually a sticker with the hardware address on it, and on a lot of routers it's external. I could have fired up the other routers and looked through their admin interfaces, but one wasn't booting anyway, and I was mad at the other... I instead did a quick poll and found the last unused IP in my block, set that, and the Internet popped on!

In a couple of days the ISP address filter will clear, and I can again use the other IPs. Of course, I can also call them or probably send them an e-mail and have it cleared manually, but I'm not anticipating rearranging anything for a few more days anyway. I've got to return the Netgear to the shop, and then find one more fully that meets my needs, probably on-line, and that has stellar reviews...like the WiFi does...

With the network all behaving again (I didn't power on the PS3 to test its connection, but I suspect it'll take the extension just fine, otherwise I'll just work my way behind the TV and plug it into the wall without the extension), I thought to test a few things. From the desktop, I got the expected peak speeds in and out. Still, though, when connected wirelessly from the Mac, only about 10Mb/s, sometimes 11Mb/s. I confirmed the 802.11n connection at 54Mb/s, but couldn't get speedtest.net to report faster than that. When wired, the Mac broke 25Mb/s.

I noticed also that they offered a speedtest "mini" that you could put on your own machine. I added it to my web server so I could test on the LAN and the WAN. The test pegs at 50Mb/s, which is much slower than my LAN, so all of the tests on the wire pegged the test. Hitting it wirelessly, even to the LAN interface on the server, still peaked at 10Mb/s on both the Mac and my phone. A little less on the phone, really, but I blame Flash for that; hitting speedtest.net gives me 10Mb/s on the phone.

I suspect there's some hosery on my WiFi becuse we also use other-than-N gear, and weak WEP because everything we connect to it works with that.

An interesting aside witht he speedtest mini on my server; it lets me check the throughput from whereever I am (or you are, if you try it). Its funny 'cause it's kind of reversed. Since my connection is asynchromous (as most are, these days), and in-bound is about five times as fast as out-bound, testing from the outside should show a peak download of abuot 5-6Mb/s (or lower if your in-bound connection is slower), and a peak upload of about 25Mb/s (again, lower if your out-bound is lower).

I just need to finish my "killer app" software so I can justify the 100Mb/s service they now offer in my area...alas, it's still asynchronous, with the out-bound speed of 10Mb/s. That's the direction I'll really need, so I'll probably have to look at co-location. Cart, horse, and all that...I still gotta finish some software...