I agree with you, though I must say, having the M.2 form factor is quite nice.
1 Like
Nice as in "cute" or "useful"?
Nice, as in 'useful.' M.2 is everywhere and dirt cheap.
1 Like
Thank you for pointing me to the what the use case for the 2.5 GbE network is! 
I must admit somehow I missed that post
What I meant is enough CPU, RAM and storage to be able to run OpenWRT or PfSense / OPNSense in their default configuration with extra capacity for use cases like @fakemanhk mentioned.
a music/video library
How much effort did you put into getting together that library?
Do you have another copy on your computer?
That is what I call a "Manual RAID 1". How much effort are you putting to keep the libraries in sync?
Here it the recommendation from Mr. Jeff Geerling himself:
for logs you really need to have replaceble storage
As far as know, the OpenWRT logs are not written to ram, are they? If you are really worried about write cycles then save stuff to some slow storage like an USB drive.
As a side note someone will want to run PfSense / OPNSense on it, although I prefer to keep things simple and use OpenWRT across everything. I would rather put in the effort to be do one thing well than two poorly.
I'm certain my thoughts will be lost in the crowd here. But I may as well record them.
Note that this is non-expert opinion, absolutely not backed up by research, and probably not worth the time of actual experts to read. So that's alright then.
First of all, I think having a standard/canonical hardware platform is an excellent idea.
Second, I'm glad that people are not being over-ambitious. Perfection is the enemy of 'good-enough'. There are a lot of suggestions for well-meaning changes, but if you are going to meet a price-point, preventing specification inflation is, unfortunately, necessary. The current specification covers many people's use-cases, where 'many' probably means 'enough'.
Thirdly, I think there is room for further iterations of the concept, which I'd like to expand.
Leaving the current specification 'as-is', the received opinion on needing to expand the capabilities of an OpenWrt wireless router appears to be:
a) Connect a switch to one of the interfaces to expand local LAN connectivity
b) Connect separate wireless access points via LAN connections to increase coverage or provide for WiFi frequencies not available in the original equipment.
c) Add a separate 'server' device to act as a NAS
d) Add a separate firewall if/when it becomes necessary
I'll suggest that two further hardware devices are needed for future capabilities.
𝛼) A box without WiFi capabilities. It should have two LAN interfaces, USB3 (no WiFi to interfere with), PCI and support for NVMe, and be as near to unbrickable as possible. Such a box can be used as an access router, a firewall, and a small server. It should be capable of running cake at the LAN interface line speed, full-duplex. I know people will want 2.5 Gbit/s LAN connections, but GigE should be fine, unless, of course, 2.5GE adds no cost disadvantage.
𝛽) A WiFi access point, capable of being powered by PoE, with two LAN interfaces that can be aggregated (if necessary - or one used for an upstream connection and one for a downstream connection to a switch), no USB3 (to minimise interference issues). It can be used to extend WiFi coverage, or act as a WiFi bridge to other devices. Also 'unbrickable'.
OpenWrt can endorse existing PoE-capable switches that can be flashed with OpenWrt.
With these building blocks, you can build a solution to meet many different needs.
I am not saying that an all-in-one solution is not necessary, or important. It obviously is the first priority. But my belief is that providing the two 'building blocks' described in (𝛼) and (𝛽) above, plus a PoE switch, would address many people's need for extra capabilities when they arise, rather than looking for a new 'perfect' single box.
I would buy a 1 Gbit/s capable OpenWrt router-without-WiFi, a couple of openWrt PoE switches and a couple of openWrt PoE-powered access points at the suggested price-point today if I could.
(Note: edited to expand disclaimer at beginning to point out my opinions are pretty worthless)
1 Like
The logs are exactly written to a ram disk to conserve the built in storage.... and I also use the external disk for persistent logging...
1 Like
First item
𝛼) A box without WiFi capabilities. It should have two LAN interfaces, USB3 (no WiFi to interfere with), PCI and support for NVMe, and be as near to unbrickable as possible. Such a box can be used as an access router, a firewall, and a small server. It should be capable of running cake at the LAN interface line speed, full-duplex. I know people will want 2.5 Gbit/s LAN connections, but GigE should be fine, unless, of course, 2.5GE adds no cost disadvantage.
Here are a few examples I can already find, I am sure there are more:
- NanoPI R6S
- 2x2.5 GbE + 1x1 GbE
- USB
-
NanoPC-T6 - Note: not yet supported by OpenWRT
- 2x2.5 GbE
- 2 x m.2 slots, 1 x miniPCIe (non LTS)
- USB
-
- 2 x 1 GbE
- USB
Un-bricking these is not as nice as the Devs device, but it's not terrible...
Second item
𝛽) A WiFi access point, capable of being powered by PoE, with two LAN interfaces that can be aggregated (if necessary - or one used for an upstream connection and one for a downstream connection to a switch), no USB3 (to minimise interference issues). It can be used to extend WiFi coverage, or act as a WiFi bridge to other devices. Also 'unbrickable'.
Do you really want to daisy chain APs? This will work once: Router -> AP -> AP .... but are you sure you want to add a 3rd or 4th one? If you can run 1 cable to the first AP, just run the 2 or 3 cables at the same for the second and third APs back to the main switch. Stay as close to a STAR network topology as you can reasonably get. If you must then put the switch before the AP, not the other way around. Same exact result, simpler setup.
I look forward to the basic. I don't want a router that is also 10 other things, I just want a well supported AP with better Wi-Fi for my family and ever increasing count of smart devices. Everything is WiFi these days, so SQM doesn't even matter as long as there is airtime fairness because I don't see being able to saturate gigabit.
Although gigabit SQM seems like a reasonable stretch goal for a 2024 device. Even though Wi-Fi 6e and 7: would be nice, if the first release is just WiFi 6 and is cheap it doesn't matter. I'll buy a new one with Wi-Fi 7 in a few years once I actually own devices that can do it.
I look forward to being able to tell people with bad wifi to just go buy this!
The 2-port version could be a base variant without wifi - for people, who love the dedicated device philosophy and like to add their own equipment (e.g. switches and APs).
The 5-port version could be an extented variant with wifi and some other nice add-ons - for people, who love the all-in-one devices.
For both variants you'll have no extra developement cost, because the base variant is just a subset of the extended variant (modularity).
5 Likes
Congrats on 20 years! This is a very exciting way to celebrate IMO. Yes, it's not the most powerful option out there, but I think it was wise to go with something that will be modestly priced and broadly appealing for their first hardware release. If this does well, that opens the door for more advanced future variants.
Also, if there was any doubt before, this confirms (for me anyway) that Mediatek Filogic is the most important platform of the present and future for OpenWrt development. Even if you don't buy this device, I think it would be wise to buy one based on similar hardware.
The BSD support on hardware is very limited, especially WiFi (5GHz WiFi support up to 802.11n only!), however now the project here is to make all-in-one so....you should probably just skip those *Sense thing.
It's in memory, not on storage (otherwise most routers die soon), but you can use remote syslog to bring it to centralized log server.
1 Like
is there a chance OpenWrt would drop charge-free iso policy and start a some payment model? my concern is OpenWrt become money, like own devices it first but huge leap to leaving free and opensource model, if someone would tell me i'm an dumb
remember redhat, remember manjaro, remember vyatta, this all started from pre-installed software diveces i mean due to self-OpenWrt device will shourly get "extended support", while all other targets would slowly die?
would be here a charge-free fork since now?
1 Like
just fork and call it LEDE 
Happy birthday, and looking forward to the One.
1 Like
so am i right? OpenWrt would split to free/paid model as redhat did?
i mean due to self-OpenWrt device will shourly get "extended support", while all other targets would slowly die?
1 Like
Very unlikely, as there are too many usecases for a one-size-fits-all device. Look at Homeassistant for a good example.
Also, too many people that are into routers know how to code for this project to just die.
/Just a random OpenWRT user.
Actually it's 20
1 Like
Congrats on 20 years ! 
On the One device, i only wish it would have m.2 2280 nvme drive because it has bigger sizes than 2tb and is more accessible, ppl likely have these more than a 2242 at hand, currently im using a Mochabin but with a 1tb sata III m.2 drive for lancaching games, though a nvme looks better anyway (removes my bottleneck), but i wish i could find something with 2280.
Of course one would say pick a server, but im just living small 
For the single 2.5G port this is for me good enough because for lancaching i often only need one port.
For the price im not complaining at all, so im considering it 
Will be there some case too? 
I will purchase OpenWrt One to show my support.
2 Likes
I will purchase OpenWrt One to show my support.
Thank you for reading and replying.
The NanoPi R6S is not currently supported by OpenWrt. It will take 'a while' until OpenWrt is on a kernel version that has the necessary support for this device. It certainly looks like nice hardware, but it's not (currently) an OpenWrt device.
The NanoPC-T6 is not (currently) supported by OpenWrt.
The NanoPi R4S is interesting, but for me, I didn't like the lack of Real Time Clock without a hack to provide the power (standard battery doesn't fit inside case), and processors are A72-based which can't do cake on full-duplex 1 Gbit/s ( https://wiki.stoplagging.com/books/technical-guides/page/nanopi-r6s-r4s-for-gigabit-sqm-with-openwrt ). Also, non-'Enterprise' versions do not have 'burnt-in' MAC address (which is important for some people).
There are plenty of devices out there that are 'almost' there, and I'm sure things will improve. Thank you again for the suggestions.
As for the chained AP scenario, I must apologise for not making myself clear. I am not proposing a chain of APs. What I was thinking was:
Upstream ISP connection - non-WiFi router - PoE capable switch - separate, individual, links to separate APs providing coverage (e.g. separate floors of house or rooms separated by radio-opaque walls). The reference to downstream of the AP was if you wanted to have LAN connections local to the AP, rather than running lots of LAN cables from the PoE-capable switch. So it would be your STAR topology.
Thanks again for the reply.