I take it to expand to "buys TP-LINK or the like, installs lede/openwrt, and ..." or are you claiming that the OEM firmwares are sufficient for most users? IMHO only a few router manufactureres/resellers actually do a good job offering updates and fixes for their already rolled-out devices. (I believe AVM does that, as does evenrouter, as well as the turris team, but tp-link IMHO is no good example for firmware updates). Now most people do not realize this (and why should they), but a router without a tested update path and an active upstream that is conscientious about timely security updates is basically just an hostile-takeover waiting to happen
I see your point, but I do believe that the all in one devices in the 50-70 $ range certainly are attractive even to this crowd (after all I might be happy to spend the time to help my family to switch to lede, but not if this costs 300$ per site for router and AP). So for a number of use-cases even the lowly MIPSen seem to be good-enough. I have a hunch that we actually do not disagree here, except maybe where exactly we draw the good-enough line
Ah, yes that answers my first point Now, it would be nice if more of the OEMs would take it upon themselves to offer devices that offer a nice and solder-free path from the OEM's firmware to openwrt/lede (and back).
But back to the topic at hand, personally I am not convinced that aiming for one reference platform would be that helpful to openwrt/lede, as I see a lot of the value in being able to convert a large number of effectively orphaned (from a security perspective) devices and convert them to openwrt/lede thereby giving them a second life instead of relegating them to the trash bin. Concentrating on a reference design has the potential to siphon resources away from the current approach of (potentially) supporting every divece under the sun.
Yes installing LEDE is what I meant. OEM is usually .... well security is the biggest issue.
For LEDE I do think it makes sense to focus on trying to support very well a few extremely common devices, and trying to support reasonably well the rest. So the C7 is a good example, lots of units were sold, and lots of people flash LEDE, and so it should work as well as can be feasibly done.
I certainly think that $300 is more than is needed. The espressobin is around $50, maybe $80 with power supply and a case? Or certainly could be that price. And a TP-Link "enterprise" access point is like $70 so we're talking maybe $150 which is the same range as the kind of all-in-one router that has similar CPU performance as the espressobin... Plus the upgrade / maintenance path is probably cheaper. add another wifi AP etc. I think it's usually the wifi chipsets and their associated power requirements that result in "burn out" on the consumer stuff. The espressobin or similar should be more reliable, and certainly less brickable with removable SD card and soforth (a good lockable case would solve any concern about an SD card with legs / security).
Well, even 150$ is on the high side for my taste; but this post made me revisit my greater families equipment, and it turns out they are all on VDSL-links using avm modem-routers (sold under the fritz-box label, being from berlin and all); these actually are not terible, at least that manufacturer has a relative decent track record with supplying devices with new formwares that offer both security updates and sometimes even new features for old devices. They also have a usable list of the specific devices support status; in short personally I like a few of the features that lede/openwrt offers, but for my extended family (which live a > 5 hours drive away) I am actually satisfied with their choice of commercial router.
The rpi is USB2-limited, i.e. can only go up to AC1200. The configuration with antennas etc. is untested and unproven. And finally when you add up all the costs, you can get close to $80 (power supply, SD, case, WiFi, etc.). That's the same retail as an Archer C7 AC1750.
Who can guaranteee though that the Archer C7 v5 or v6 won't be Broadcom-based? And then you need to find another device. Same happened with the WRT54G and so many other h/w platforms and Wifi chipsets/drivers. Think an evolving reference architecture would be of great benefit besides all the supported devices. Why not one more that you control better and which would synthesize all the h/w learnings?
No, the WRT54G became dated in its technology; compute power, memory, and eventually wireless standards.
Yes, poor decision on their part for the router market, yet no reason to add more expensive components for their market. ODRIOD "work-alike" units aren't, and I've run well over 500 Mbps through their on-board Ethernet adapter and a USB-3 Ethernet adapter.
Because there are far too few OS and application-code experts willing to contribute their time, and I suspect even fewer hardware experts.
The forestâ could care less â theyâre going to buy whatever their service provider offers them, or Amazon or Walmart or the like has cheap...
Is there an opportunity to show the benefits of an openwrt router? Not the "you can configure it to do complicated things" benefits enthusiasts like, just simply it is faster. Show off graphs showing improved ping in situations when buffer bloat is a problem for gamers and VoIP users. Show that with many old devices attached it still provides great throughput for watching 4k Netflix (thanks to recent airtime fairness work). Users don't need to care this is thanks to openwrt/open Linux drivers. They just need to know their old router 5300ac router isn't scaling up to their household needs and this is better suited, even if it was only 1200ac!
It wouldn't be specifically openwrt related because routers with closed drivers like Marvell chipsets don't get all the benefits mt76 based routers do. To the end users it would simply be a router they could plug in and scale well to alo their devices.
Don't know if you remember when Cisco acquired Linksys and the WRT54G got half of the RAM because of cost cutting (think it got down to 8MB from 16MB). At some point later they released the WRT54GL version but that did not continue, it was just to please the community. This was just an example of how generations of good h/w platforms have been discontinued because the economics at some point shift. What if the merger of Broadcom and Qualcomm had gone through? What would have happened to the FLOSS drivers? Yesterday the "star" router was Linksys, today TP-Link, tomorrow someone else.
Because there are far too few OS and application-code experts willing to contribute their time, and I suspect even fewer hardware experts.
That does not make sense. Why would they be contributing their time today on a random long tail of devices that may or may not have future Vs. one that they will have designed and which will have the community's support behind it.
It is not that I do not see the charm of your idea, it is more that I do not believe this to be realistically achievable; or rather just get a turris omnia that is pretty close to your requirements, except for the price (but that is going to be an issue with any special tailored device, if we want commodity pricing we need to accept commodity market practices). Sorry to appear contrarian here...
Well except it seems maybe more expensive than an x86 from xcy plus a separate access point.
I think if you want really cheap you have to be flexible in choice of components, hence the constant revisions by tp link etc. If you want fixed components that are well supported you need to pay substantially more, and by the time you're doing that, might as well go with the low end x86 which is well supported and has a ton of horsepower.
Agree on the ARM-based approach, but do not think it's feasible under $80-90.
And I'm picking the $80-90 range since this is the sweet spot where the Archer C7 plays and it's become the most popular router.
If you go above, I think you're trading off between a good WiFi chip and the CPU architecture.
You need to go $150+ to get the best of both worlds.
this post made me revisit my greater families equipment, and it turns out they are all on VDSL-links using avm modem-routers (sold under the fritz-box label, being from berlin and all); these actually are not terible, at least that manufacturer has a relative decent track record with supplying devices with new formwares that offer both security updates and sometimes even new features for old devices. They also have a usable list of the specific devices support status; in short personally I like a few of the features that lede/openwrt offers, but for my extended family (which live a > 5 hours drive away)
AVM is a great vendor, they do support their devices over an extended period of time. However, I do not subscribe to the notion that hardware has to be bundled with software, particularly when the economic incentives are tied to the hardware, i.e. you make money based on spec-based sales and not software-driven value. This gives you the wrong set of incentives. You're as a vendor incentivized to switch out any subcomponent to improve your profit margin and you're not capability oriented. If they were selling software, this would be different. All the box moving vendors are motivated by what I just described. That is why, I think, LEDE/OpenWRT from the software side could be dictating what they would want with a reference architecture roadmap, building their own portal and letting people buy their designated equipment from 2-3 OEM vendors who they would keep a relationship with, much like to what the raspberry pi does.
The Qualcomm IPQ40x8/9 - Dual-Radio Networking WiSoC could also provide a decent alternative for a low cost router that could be used with ath10k. Otherwise Mediatek seems to be the way when it comes to low cost AC routers. The Asus AC-RT58U is a good example in this category. Any views on ideal Mediatek-based devices? Is there a cost-performance/functionality tradeoff when thinking between Qualcomm and Mediatek?
A lot of people who know could strongly argue in favor of ath10k Vs. mt76.
Could you explain as to why the DIR-860 is the most interesting router? Why would it be better than an IPQ-based solution with ARM architecture?
I slightly disagree, avm sells hard- and software, it is partly due to their reputation of relative long-term support for their devices that they are as popular as they are over here, so the software side make the hardware sell well. The bundling also allows them to sell something tangible instead of "just" software. Software support will either need to be temporally limited or sold as a recurring "lease" otherwise there is not sustainable way of supporting a company, by selling hardware with a limited guarantee for software updates AVM avoids that pitfall (and they are also good sports in sometimes even supplying security updates for devices that are officially not supported anymore*).
It is exactly this cut-throat effect of the market in combination with the requirement to up the advertised specs every now and then that keeps the commodization wheel going, otherwise parts would not get cheaper or better that fast; will this result in "ideal performance" or "best quality", probably not, but "good enough" certainly seems achievable. But you need volume to drive down prices...
As stated above, I disagree, I do not believe that the market is ripe for a software-only vendor (especially since openenwrt/lede is free and the most recent security updates have been quite timely there is only little air left for commercial software services on top of openwrt).
I am not opposed to that, I simply see no driving force that would make that happen. And I actually bought a turris omnia whch promises a lot of the things you seem to want. And I like the TO a lot, but still my main router runs openwrt master...
*) One could argue that this is not only customer friendly but also economically sane to fix their own exploitable bugs independent of support periods to avoid being sued, but still there certainly are companies that will go the sue me fist route, so kudos to AVM.
X86, arm, MIPS, doesn't matter when you've got source instead of legacy applications. You can look at the performance and decide if it is high enough or not. I believe the 860 has a quick dual core CPU that will be faster than the c7, if you ever are in a spot that matters.
The best thing the ath10k has is its name. It brings up memories of the open ath9k. It's fine hardware, but the drivers aren't open to build the best openwrt software like the ath9k was. It's closed like the the Marvell chipset.
This matters if you care about being open. You care about being open if you want support down the road or if you want cool new features like the airtime fairness work. https://arxiv.org/pdf/1703.00064 "The airtime fairness scheduler implementation is limited to the ath9k driver, as the ath10k driver lacks the required scheduling hooks."
If you're looking for an ideal router why would you choose one that only uses the subset of openwrt with no chance of getting improvements like the airtime fairness scheduler?
Except that high performance MIPS "died" with SGI switching to Itanium, and ARM is still trying to establish that it really can x86 when talking about pure CPU oomph; POWER probably is a worthy contender, but unlike low- to mid-tier x86 power is not an option for home routers...
Well, I think ath10K also has candela tech going for it who signed the necessary NDAs to be able to roll their own firmwares, so surely not as open as 9k still better than others...
This is partly caused by ath10k simply doing way more inside the firmware than ath9k, I believe there still is an open ath10k driver, but that simply does not expose/know enough about the low level stuff to be useful for Toke's airtime fairness approach. (I do hope that @greearb will be able to bridge the gap to the hardware, but that might be a bit optimistic).
I am not sure that the ediatek story really is that much better, but am willing to be convinced otherwise, but typically the interaction between people who sell their hardware with matched SDKs and the open source folkes is best described as "it's complicated"...
