Netgear now touting open source WRT-compatible wireless router

Ben Scott dragonhawk at gmail.com
Tue Jul 1 11:18:17 EDT 2008


On Mon, Jun 30, 2008 at 5:14 PM, Arc Riley <arcriley at gmail.com> wrote:
> If it was really designed for hacking it'd include more flash, ram, and
> pinouts for expanding/hacking the hardware.

  "Designed for hacking" is relative.  When LinkSys, NetGear, et. al.,
say that, what they mean is:

A1. It has enough computrons to run a reasonable Linux install.
A2. They didn't go out of their way to prevent third-party firmware
from running on it.

  They're not talking about hardware hacking.

  It may be interesting to speculate as to why.  Here's my reasoning:

  Current versions of the LinkSys WRT54G-without-the-L series no
longer have enough storage to run a reasonable Linux kit.  (They can
boot Linux, but only a severely feature-constrained variant.)  LinkSys
switched to the smaller-footprint VxWorks to let them ship the
WRT54G-without-the-L with less RAM and less flash ROM.  That let them
knock a few bucks off the BOM cost.  For the product space they're
working in, that's huge.  BOM cost is the overriding concern for all
design decisions.  Everything else is secondary.  *Everything*.  Until
you've talked to engineers working in this product space, it's hard to
appreciate the cost pressure they work under.  They'll gladly redesign
a whole PCB to shave off the cost of a 50 cent part and the total
manufacturing cost it adds.

  So, I presume that the WRT54GL line was introduced because:

B1. It turned out that a lot of people were buying the WRT54G to run
third-party firmware.
B2. As a consequence of B1, older WRT54G revs started selling for much
higher than MSRP on eBay, et. al.
B3. It is relatively cheap to switch a manufacturing run to larger
capacity ICs on an existing PCB design.

  LinkSys saw a market opportunity.  They could offset the cost of B3
while still beating out B2 (and making a tidy profit on the side, I'm
sure).

  But changing the PCB design (e.g., to add I2C headers) is a bigger
step.  Not much, but it means another part in inventory, another part
during manufacturing, another thing to test after manufacturing, etc.
With the razor-thin margins and large unit volumes in this product
space, that can be more significant than one might expect.

  That said, if this was just about total manufacturing costs, I'd
guess they'd be able to charge an additional $5-$10 without loosing
significant sales.  They're already targeting a much narrower market
segment, and one which has demonstrated a willingness to pay a price
premium.  That additional money could pay for things like adding an
I2C header.

  *However*, LinkSys, et. al., are really not supporting third-party
modifications.  Once you load third-party firmware, the warranty is
void, until and unless you re-load the OEM firmware.  If you can't
reproduce the problem with the OEM firmware, too bad.  And if the
functionality you're trying for doesn't exist with the OEM firmware,
too bad.  If they started advertising pinouts and putting in
additional connectors, they'd take on a support burden they have no
interest in.  (If you've ever contacted LinkSys support, you'll know
they can barely support the stuff they have now.)

  And no, putting on the connectors but saying they're not supported
is not a viable solution.  For one, in many jurisdictions, there's an
implied warranty that the product is fit for the intended purpose.
For another, if it's there, people will call support about it, even if
you print in inch-high red letters that it's not supported.  LinkSys
doesn't want to absorb that cost.  (See above about razor-thin
margins.)

  It does occur to me that there might still be an opportunity here
for LinkSys (or whoever): Manufacturing base boards for
value-added-manufacturers (VAMs).  If LinkSys ran traces but just left
bare pads for certain things, VAMs could buy them in bulk, add on
additional features, take on the customer support burden, and make a
profit reselling them.  But I'm guessing LinkSys isn't prepared to
even support VAMs in this way.  (Again, they can barely support their
products when used as intended.  This stuff is cheap first, everything
else second.)  Not to mention the mindset change it would require.
Doing something other than "dirt cheap" is really alien to these
people.

  So, if one wants customizable hardware that actually works, I'd
recommend buying something designed for the purpose.  Soekris
(http://www.soekris.com/) and Koolu (http://www.koolu.com/) have
gotten mention on this list as working well for this sort of thing.
They make bitty-boxes with modular RAM, ROM, I/O, etc.  Of course,
they cost more than the SOHO routers you find at Wal-Mart.  TANSTAAFL.

-- Ben


More information about the gnhlug-discuss mailing list