We all know the saying about the weakest link and recognize its inherent truth. In the great chain of technology that is our broadband service, we seldom think of our state-of-the-art routers as being that link, but perhaps we should.
Recently I wrote about the mystery of how one can never seem to get, say, 3Mbps out a 3Mbps broadband circuit. One reader responded with a tale describing his extensive efforts to coax more than 860Kbps out of his broadband. With no router -- just a direct-connect single-station connection -- he got 1.5Mbps. Ultimately, he wrote, it was the "molasses factor" in the router innards that caused the dramatic degradation of throughput.
When you walk through your local IT retailer in search of the perfect broadband router, I doubt the expectation of submegabit-per-second throughput would even cross your mind. After all, broadband routers (in places like the US) have been sporting 100Mbps (Fast Ethernet) interfaces for years now, and even Gigabit interfaces are increasingly common. It is a long way down from a billion bits per second to less than a million bits per second. So what gives?
To a large extent, the problem might be the disposable, pack-of-chewing-gum approach that some vendors take toward building and supporting the boxes. Many vendors have so many products you can't imagine them spending much time on any given one. You probably spent more time on your Christmas card list than some vendors spend on bringing a broadband router to market.
Pick a site, Linksys, for example, and click over to the support downloads menu. There you will see a hundred or more products and versions listed. One product I picked at random showed about 45 "fixed problem" comments and only five "added support" comments in the two-year life of the driver. The impression, at least, is one of quantity rather than quality: Throw it out there, fix the biggest problems and get the next box out.
With our reader (who wasn't working with a Linksys product), it took the next box to resolve his problem and allow a blistering 1.5Mbps of throughput. He clearly spent much more -- in terms of his own time -- than the box cost. Does the vendor care? The reader wonders, as I do in similar situations, how nontechnical persons would deal with the issue. They couldn't.
Keep in mind how many broadband routers come into being. It almost is akin to working with a Lego set - or maybe ordering up a McDonald's Happy Meal for your kid. You can order up, say, Broadcom-based hardware and Jungo's OpenRG development platform, and you've pretty much got yourself a router faster than you can order up a root beer.
Not to say that vendors routinely slap things together or that there is any inherent problem with the aforementioned components, but we've all seen how less than optimal code can result from the level of prebuilt code used and vendors' desire to keep build costs as low as possible.
What to do? When you get a router, test it out. When you find a good one, keep it. That shiny new one you are thinking of buying could be a step backward.