Netgear's open router

Your editor was recently reminiscing about an early stage of his career, which involved the administration of a VAX 11/780 computer. The VAX was a highly successful product, as was its native operating system VMS. Quite a few VAX customers chose to do without VMS, though, and put early versions of BSD Unix on them instead. Digital Equipment Corporation never entirely appreciated those customers. To DEC, every BSD installation looked like a lost VMS service contract.

The company should, instead, have seen those installations as an extra sale gained as a result of the VAX's ability to run a nice operating system.

Almost 30 years later, some parts of the computing industry have come to understand that there is value in selling hardware which can run operating systems provided by others. Microsoft made that point in a big way, of course, but there are also significant parts of the industry which benefit from making systems which can run Linux - and, in particular, a version of Linux which is not necessarily supplied by the vendor.

But other sectors still seem to see the ability for the customer to put (or replace) Linux on their systems the way DEC saw Unix in the early 1980's. They see no value in letting their customers make changes to their systems, choosing instead to lock those systems down and keep total control. Embedded systems are often singled out as an example of this type of behavior, and vendors of small routers tend to be especially inclined in this way. It is not a coincidence that a substantial portion of the high-profile GPL-enforcement cases to date have involved consumer-level routers.

Some vendors, at least, are getting smarter and doing what they need to do to avoid licensing problems. But relatively few of them welcome customers who want to replace the software on "their" devices. There are exceptions, though, and their number just grew with this announcement from Netgear. The WGR614L router looks like a fairly straightforward consumer wireless router, with the usual set of features. LWN readers will doubtless be glad to hear that it is "Works with Windows Vista" certified. It has a four-port Ethernet switch, an 802.11g access point, and a mighty 240 MHz CPU and 16MB of RAM. All of the stuff one would expect from an inexpensive desktop device.

But what makes this device interesting is that it's designed to be open and hackable. The source code for the factory-installed firmware is available from Netgear's community web site; it's amusingly packaged as a zip file containing a single, compressed tarball which, in turn, holds a bleeding-edge 2.4.20 kernel tree. But anybody wanting something a bit more contemporary and community-oriented can replace that firmware altogether with a package like Tomato or DD-WRT; indeed, Netgear almost seems to encourage its customers to do so.

Every one of those customers then gets the benefit of the effort which has gone into the development of those router distributions - with little effort required on Netgear's part. Those customers can improve this platform and make their changes available to other customers; that makes Netgear's hardware more valuable. If there are bugs in the system, a single motivated customer can fix them and make those fixes available to everybody else. And all of this comes at almost no cost to Netgear.

It is always fun to see Linux turn up in new places. It's now a routine experience to realize that one's new television, camcorder, music player, or automobile runs Linux. But locked-down, Linux-based devices are not far removed from the fully proprietary systems which preceded them. Whether or not one agrees that locking down systems in this way is legally or morally defensible, it's easy to conclude that it is undesirable. A Linux system which is cast in concrete loses a part of the vital energy which makes Linux what it is.

So it is always a welcome development when a vendor decides to take a more open path. With any luck at all, the wider public will eventually realize that more open devices are more powerful devices, and, as a result, such devices will prove more successful. That is the path that brings us more control over our systems and, eventually, to World Domination.

---

(Log in to post comments)

Alas, Netgear based this router on a Broadcom chip (BCM5354) for which there is only a
closed-source wireless driver. The same problem has hobbled development on the venerable
WRT54G for quite some time, precluding projects like openwrt from offering up-to-date 2.6
kernels for the hardware.

Eventually (hopefully soon) the reverse-engineered, GPL-licensed b43 driver will be ready for
prime-time, but in the meantime this is an "open" platform that requires a vital closed-source
component.

For anyone looking to upgrade their router, it's worrying that there are currently no
open-source drivers for any (pre-)802.11n chip. From the sidelines, it seems ath5k may get
there first, but the impression I get is that Linux is losing out to closed-source solutions
in the world of embedded code for routers.

I'd be interested to hear comments from any openwrt developers who are lurking in these parts.

Do realtek chips show up in routers? They've had 11n support in their drivers for some time,
even if it isn't in the kernel yet.

There are a few that have ralink. Most have Atheros and Broadcom though. 

What may be worthwhile is to get a device that uses mini-pci slots, then you can put in
whatever you want.

And there are a few 'high end' or 'professional' versions for people making their own things
for custom solutions. They'll sell ARM boards and such with no cases or anything, but have
mini-pci slots.

If your curious then check out their "Table of Hardware"
http://wiki.openwrt.org/TableOfHardware

> For anyone looking to upgrade their router, it's worrying that there are currently no open
source drivers for any (pre-)802.11n chip.

I might be missing something, but the Intel pre-n card in my laptop works fine. No idea
whether it works in -n mode though. Maybe that's what you meant...

Intel's work on free wireless drivers is great (I'm typing this on a machine with an IPW2200),
but last I looked their driver for the 4965AGN pre-N chip didn't support AP mode.

Apart from that, I don't believe there's a low-cost router platform using Intel wireless.

None of the current mac80211 drivers in mainline support AP mode at the moment. But the work
that needs to be done for this is mainly in the mac80211 layer and in hostapd, not the drivers
themselves.

Once this work is completed, the majority of wireless devices will gain AP support all at
once. However, don't hold your breath. It hasn't been a development priority in the last few
months.

The tone of this article gives the impression this router is the first of its kind. Linksys
has been releasing the source of the GPL parts of their firmwares since ages : after some
pressure (apparently) they released the source of their WRT54G router in 2003 !

So while it is nice to see another manufacturer playing by the rules, it is nothing new in
this area.

Unfortunately the WRT54G(L) routers also use the Broadcom wifi chipset, so you're stuck on the
2.4 kernel until the b43 driver matures. That fact only reinforces your point that this isn't
another first. Good to have more manufacturers on board though.

Not to mention that as far as I'm aware, Vista does not currently run on
the MIPS32 architecture. I think it's safe to assume the "will connect to
machines running Vista" interpretation.

I believe that means it implements the "Universal Plug & Play" interface, maybe with some
extensions added for Vista.

Anyway, that is the first thing I disable on any router I can. The security aspect of having
applications enable port forwarding remotely seems to me one of the stupidest things ever
made.

It would be great if it was actually current spec hardware with full
open source drivers.

I would love to see a fully supported home router with ADSL2+ and 802.11n draft.

Unfortunately openwrt.org and other projects are falling behind...

You mean the manufacturers making such routers, where free-software based, are falling behind
in providing the required source...

Infineon have a DSL router chipset with DSL2+, I think. They've provided resources to OpenWRT
developers to help get those supported, AFAIK.

Linksys is pretty good about publishing source code:

ftp://ftp.linksys.com/opensourcecode/

Although they do include binary-only drivers as well.

What about seeing our beloved operating system kernel included on the embedded systems found
in our tube transportation system, our plants' control system, our planes' avionics systems,
our voting machines, our satellites' ground stations?
 Would you really like to see them as hackable as the Netgear's router?

NB: This is a real question. Honest! I have no clear position on this up to now personnally.

The criteron for hackability in this case is that the people who have the 
rights to use the device should have the rights to modify how it works.

In an avionics system or a voting machine or a metro system, the only sane 
definition of `user' is `the airline / election authority / metro 
company'. They shouldn't be stuck with a bunch of closed stuff they can't 
hack at, validate, and so on (and avionics companies insist on this: I 
wish election authorities did).

However, equally, I don't have anything but an intuition that this must be 
so: I can't formalize it. Anyone who can, please chip in.

I see this announcement as too little, too late.

This effort is essentially decoy PR so Penguinistas don't flock to another, more open router
platform.

They need to open up the draft-N routers in order to be credible.  I tried working with the
company to make that happen, but I hit a brick wall rather quickly.  

Because they obviously didn't care, a while ago, I decided to boycott Netgear due to this, and
have actively discouraged customers and others I have a relationship with from buying their
equipment.  (Not that it matters, mind you).

When they have a hackable N platform, then I'll believe they're serious.

When they have a hackable platform in a nice steel box, *I'll* believe they're serious.

The 524/614/624's are in that "Dies 13 months after you buy it" club, in my experience.

240 MHz CPU and 16MB of RAM should be plenty enough to run VMS on that :>