Brainstorming: Routers without a Firmware , Why Not ?

[IAAI]

I m thinking of a solution could make everyone happy . What would happen if for example xyz produce a router without a software ? Think of it like a regular PC without OS. You choose what operating system you like the most .

*Routers would be at least 30% cheaper - The Hardware game begins

*No software headache for the producer - They will race to make their hardware compatible with xyz fw company .

*Opening up wild gates to software Companies & Professional Networking engineers (Who Really Listen To US) attracting customers to buy their FW .


------------
what is your opinion on this idea ?
disadvantages ?
advantages ?

 

[azazel1024]

Wouldn't work. You must have some firmware to begin with, even if it is only a simple boot loader.

Otherwise it would be more akin to shipping a PC with no BIOS/UEFI loaded on it. Good luck with that.

SOMEONE has to develop firmware for it. Open source doesn't necessarily work and it certainly adds significant time to market.

"Hey guys, buy our router. Its some hardware and...uh...feel free to wait a few months till someone gets one and develops firmware that'll work on it".

The firmware that runs on them is (/needs to be) custom enough that you can't simply load any build of DD-WRT or whatever on it. It is not as general purpose as, for example, Windows or Linux (full linux). If you wanted that kind of interoperability without custom builds, you'd likely be looking at hundreds of MBs for the firmware install, plus more processor overhead and RAM overhead, even if it was just loading the specific API, instructions, etc required for the hardware present.

Its unlikely to happen any time soon. Plus the manufactuer has to have firmware to begin with to test things like the hardware. So why not "finish" the firmware through a full release instead of leaving it in an alpha state.

 

[TonyH]

Hi,
Of course, if one can write a perfect software(firmware) and produce ROM or ASIC with it but who can write perfect any thing? Think definition of firmware.

 

[IAAI]

Wouldn't work. You must have some firmware to begin with, even if it is only a simple boot loader.



Otherwise it would be more akin to shipping a PC with no BIOS/UEFI loaded on it. Good luck with that.



SOMEONE has to develop firmware for it. Open source doesn't necessarily work and it certainly adds significant time to market.



"Hey guys, buy our router. Its some hardware and...uh...feel free to wait a few months till someone gets one and develops firmware that'll work on it".



The firmware that runs on them is (/needs to be) custom enough that you can't simply load any build of DD-WRT or whatever on it. It is not as general purpose as, for example, Windows or Linux (full linux). If you wanted that kind of interoperability without custom builds, you'd likely be looking at hundreds of MBs for the firmware install, plus more processor overhead and RAM overhead, even if it was just loading the specific API, instructions, etc required for the hardware present.



Its unlikely to happen any time soon. Plus the manufactuer has to have firmware to begin with to test things like the hardware. So why not "finish" the firmware through a full release instead of leaving it in an alpha state.

What if the router comes OS-Ready/Preloaded with xyz trial fw
As a Marketing strategy

 

[IAAI]

Hi,
Of course, if one can write a perfect software(firmware) and produce ROM or ASIC with it but who can write perfect any thing? Think definition of firmware.

They can if they want too + we are not talking about 1 but hundreds of devs their only job is to get it working probably such as (planning + developing + implementing + testing + debugging)

 

[htismaqe]

Software is a market differentiator.

Even in the coming world of SDN, vendors are positioning devices to load THEIR software, not software of the user's choice.

Vendors are not going to do away with market differentiators.

 

[RogerSC]

Not having firmware would also prevent manufacturers from putting routers on the market without all the promised features, or working in some reduced mode. This would increase the time to market, and decrease the competitive advantage that they think that user disappointment gives them *smile*.

Anyways, they need the ability to fix bugs, especially security bugs, as time goes on. If security bugs/exploits turned up after their initial release, they'd have to send you a new router to fix them *smile*, or you'd have to buy a new router to get fixes. Not that having users have to buy a new router to get security fixes would displease manufacturers, but that would be a different world.

 

[azazel1024]

What if the router comes OS-Ready/Preloaded with xyz trial fw
As a Marketing strategy

See "open source ready" routers.

Again, see what I mentioned about the issue of OS size. Things like Windows, Linux, etc are so much larger, not simply because of the features, but also the fact that they need to interoperate on so much different hardware.

Granted, we aren't talking a 3.6GB ISO to load up on your router for router firmware that would support most/all hardware out there, but it might turn in to a 20-30MB, or even larger firmware file...that then gets decompressed in to something a lot bigger. Flash would need to get bigger (more expense), RAM requirements would be higher, start-up times would be longer as you'd have to figure out what modules to load in to memory. Also likely it won't really be optimized to the hardware.

DD-WRT doesn't run on all DD-WRT capable routers. Specific builds run on specific routers, all under the heading of "DD-WRT". It's like saying "linux". There is RedHat, Debian, Ubuntu, etc, etc. Except in this case it would be hardware specific versions of Linux (which...guess what the underlaying OS is for most routers)

Its a lovely idea, but when it comes down to an appliance, which is what a typical router is, it won't really ever work. Now, if you went to SDR (Software Defined Routing), you can do it. By all means. The problem there is, again, very high requirements and it is difficult to get in to too much hardware accelerated functionality, because fixed function hardware support, across a variety of implementations takes more and more storage space to be able to account for how to work it for the OS/Firmware. SDR has the promise (and can be) awesome, but the hardware is vastly more expensive for a reason, and it isn't always faster either (slower doesn't necessarily mean it isn't better).

Just like software defined switching.

 

[RMerlin]

A router without a firmware is basically just a switch. You need code to process the packets according to user parameters to handle routing. That's what differentiate a switch from a router.

After that, anyone developping a firmware needs to control the hardware. This is what the driver code does. An operating system without driver code doesn't do anything - it needs to interface with the hardware. That's where the hardware manufacturer has to provide you (the firmware developer) with driver code to use their hardware. Writing device drivers to interface with hardware can be very time consuming, and requires very low level knowledge of the hardware - something that is not publicly available for SoC (see how much Linus Torvalds love SoC manufacturers, for a good example. Expect your usual Linus expletives. )

And you're back to your starting point: router need to ship with some code from the get go, even if you were to develop your own firmware for it.

Kinda like how Windows or Linux wouldn't do anything without the driver code to make it talk to the hardware.

 

[TonyH]

They can if they want too + we are not talking about 1 but hundreds of devs their only job is to get it working probably such as (planning + developing + implementing + testing + debugging)

Hi,
Let me ask you a question. Have you seen a perfect programmer or perfect
software(firmware)? I never did in my 40+ years in the field. I spent chunk of that 40+ years applying patches, applying engineering changes, etc. which never ends.

 

[TonyH]

What if the router comes OS-Ready/Preloaded with xyz trial fw
As a Marketing strategy

Hi,
You are talking about todays routers, No?

 

[RMerlin]

I spent chunk of that 40+ years applying patches, applying engineering changes, etc. which never ends.

And things keep getting worse and worse, as software is becoming ridiculously complex as customers expect device "X" to do everything and some more. My Amiga 1200 had a complete operating system taking a few megabytes, and the hard disk on it was 20 MB. Today, a *MOUSE* driver wouldn't fit on that 20 MB HDD...

And also as time-to-release are getting shorter. Manufacturers ship devices with known defects due to running out of time to fix it, and expecting customers to get software updates through the Internet at a later time.

After 15 years as an IT professional (and 30 years playing with computers and other devices in general), I've grown quite disillusioned about the current state of software in our world.

Someday, something BIG will crumble due to software issues.

 

[htismaqe]

And also as time-to-release are getting shorter. Manufacturers ship devices with known defects due to running out of time to fix it, and expecting customers to get software updates through the Internet at a later time.

This methodology is rampant in the video game industry.

 

[RMerlin]

This methodology is rampant in the video game industry.

Day zero patches anyone?

I also remember my Motorola Xoom. Android 3.0 had no support for the SD card slot they had added to the device. Took until the 3.1 Android update, nearly a year later, before the SD card even worked...

Also, I know someone who works at a very well known business and home software publisher. He was telling me once about when he had to ask someone else at the office: "Are you telling me we are going to ship a product with "xxx" known defects?!" (I forgot the actual number, but it wasn't just 3 or 4)

 

[htismaqe]

Day zero patches anyone?

I also remember my Motorola Xoom. Android 3.0 had no support for the SD card slot they had added to the device. Took until the 3.1 Android update, nearly a year later, before the SD card even worked...

Also, I know someone who works at a very well known business and home software publisher. He was telling me once about when he had to ask someone else at the office: "Are you telling me we are going to ship a product with "xxx" known defects?!" (I forgot the actual number, but it wasn't just 3 or 4)

It's one of the reasons so many consumer electronics companies now make devices that REQUIRE an internet connection.

 

[azazel1024]

Day zero patches anyone?

I also remember my Motorola Xoom. Android 3.0 had no support for the SD card slot they had added to the device. Took until the 3.1 Android update, nearly a year later, before the SD card even worked...

Also, I know someone who works at a very well known business and home software publisher. He was telling me once about when he had to ask someone else at the office: "Are you telling me we are going to ship a product with "xxx" known defects?!" (I forgot the actual number, but it wasn't just 3 or 4)

Similar story. I'll out the company (maybe different than yours), Microsoft. One of the medium higher ups gave a speech at our agency years back right around when I started. They were talking about their software design approach and the exec boasted that they shipped Outlook (2003) with only (WAG, as I don't remember the exact number) 213 known bugs!

Crickets were heard.

I think the most bugs I've seen in an application we've deployed is around 20, with none of them above the level of medium (on a scale of "Not a Problem", low, medium, high, severe and critical). At medium might be something like an error message doesn't display quite right, or should have been a java script message, but only appears as an in-line redball error on the page. Low would be something like a spelling mistake.

High (and above) would be things that legitimately make the application not work right, no matter how rare it might be.

Most I've seen ship with fewer than half a dozen known bugs (mostly low/medium) and sometimes zero known bugs.

Of course that doesn't mean that problems aren't found later. With a typical release cycle, we are looking at 6-10 validators who are 30-80% time on validating an application for 2-4 months. Plus a month or two of integration time (in a mirror of our production environment) and then to the public. Compared to the handful of work years total of validation time an application might get, some of the applications have several MILLION users...defects are found later. We also generally have quick turn around on fixing them, from months for minor ones when the next maintaince release is scheduled, to less than a week for any exposed critical defects.

 

[TonyH]

Hi,
Interesting is often patch to fix a bug creates another bug, LOL! More scary is drugs dealing with human lives, same is going on.... We're never perfect, forever we're trying to be perfect.

 

[sfx2000]

I m thinking of a solution could make everyone happy . What would happen if for example xyz produce a router without a software ? Think of it like a regular PC without OS. You choose what operating system you like the most .

*Routers would be at least 30% cheaper - The Hardware game begins

*No software headache for the producer - They will race to make their hardware compatible with xyz fw company .

*Opening up wild gates to software Companies & Professional Networking engineers (Who Really Listen To US) attracting customers to buy their FW .


------------
what is your opinion on this idea ?
disadvantages ?
advantages ?

Thoughts here...

When a router/AP is designed, the major costs are the PCB, the Chipsets, and the rest of the Bills of Material - software, for the most part, isn't much of a cost adder.

Walking things thru - let's say I want to build out a AC1200 class router. N600 for 2.4GHz, and AC1200 for 5GHz. Simple enough these days...

I can do two paths, and much if it depends on maturity of the technology, but also the business case...

Path 1 - I'm a big name, and it's a product that isn't out yet, but silicon is baking, and close to mass production.

I call down the road to Broadcom, Atheros, and Marvell - I tell them I want X/Y/Z - they give me a proposed chipset, gerbers for a reference design, a board support SW package, along with a couple of SDK boards (stretches). This along with a basic bill of materials for the reference design so that the Product guys can start working their numbers...

This is basic router/AP - now if I want to add a DLNA/Airplay/TimeMachine/NAS etc, then I've got to add the appropriate HW/SW needed on top of the Reference HW/SW design (this is where most bugs creep in)

Add that... And because this is new stuff, there's a fair amount of NRE costs that are going to go with it.

Path 2 - mature stuff... even if I'm a Tier One OEM, when we get into the rinse/lather/repeat mode in the business cycle, I still want to get "new" product onto the shelf - so I go to an ODM - these are companies you here about once in a while - Cameo is one, Gemstar is another - Taiwan is full of the white box folks that build things, and they're very good at it. And the margins here on Manf, they're very tight, as expected, because the Tier1 OEM is working to recover the NRE spent on the leading edge product.

There again, we talk about features, and basic costs FOB somewhere, these days for the US is typically either LongBeach or SFO/Oakland.

Really what it comes down to - this process is iterative - from 1st Gen through followups - and it's tightly coupled on the HW and SW.

To roll one's own - it's entirely possible - MicroTik is a good example of this with their routerboard line, and when you get the HW, they have a link/license for the SW, which is open source, and one can tinker at will... there are others...

in the end however, the SOC/WiFi chipset vendors - they want you locked into their platforms on a generational basis - they make the code and toolchains exclusive, and the cost to jump out very high indeed...

Qualcomm has been banging that drum for over 15 years in the mobile space, and that's why everyone else is now having a hard time breaking in - it's not just the basic BOM, it's the tools and dev environment, and that is where the lock in exists.

sfx

 

[stevech]

I m thinking of a solution could make everyone happy . What would happen if for example xyz produce a router without a software ? Think of it like a regular PC without OS. You choose what operating system you like the most .

*Routers would be at least 30% cheaper - The Hardware game begins

*No software headache for the producer - They will race to make their hardware compatible with xyz fw company .

*Opening up wild gates to software Companies & Professional Networking engineers (Who Really Listen To US) attracting customers to buy their FW .


------------
what is your opinion on this idea ?
disadvantages ?
advantages ?

Right! Concept applies to buying a car, too.
You buy without engine. You drive to engine-of-your-choice shop.

 

[4Leaf]

The software is the brains of the operation and tells the hardware how to behave. I believe the best solution is to figure out a new way of making the software. We just need to figure out a better way to present the language coding that could perhaps possibly produce way less bugs in the software.