I have at home several older devices which I can donate for OpenWRT development:
Zyxel NWA-1100-N
this one is not a router, it's an PoE AP/client/bridge CPU1:Atheros AR7242 (400 MHz) FLA1:8 MiB (Winbond W25Q64BVSFIG) RAM1:64 MiB (Winbond W9751G6JB-25)
Exactly! That doesn't make sense to stop supporting devices when there's more than enough flash space available. Devices should be supported for as long as possible.
In reality this is not doable for a lot of reason. Starting with the question what "as long possible" mean. But the main reason is that Developers move on as well as the life for everybody does. Like:
They do not own the device(s) anymore.
They have newer devices which they want to spend their time on resp. you can see the demand for new devices in the dev section here.
They don't want to put time into finiding hacky solutions for corner cases.
They don't want to deal with those old devices anyway (how selfish )
On top of that: Old devices tend to have almost no userbase anymore. This is due to changing demands (in life). E. g. Home Office with 250MBit+ internet speeds cannot be handled by any 8/64MB devices. Neither via WAN nor via Wifi (at least I don't know one from the old ones). No point to stick to devices 10y+ old like a Netgear WNDR3700 which was a great device in the past. They've advertised 1G WAN speed if I remember correct. But it was never capable in reality. Nobody was able to check this because nobody had a gigibit WAN connection in the past. ;D
There are dev's spending time getting older versions of OpenWrt running on thoses devices out of spec. But those efforts are limited in scope for various reasons (e. g. kernel size growth).
Unfortunately most manufacturers provide security updates for their routers only for couple years. So, one of the main reasons to install OpenWRT is to keep the FW secured. For example one of my family members still uses my old Linksys WRT610N (DD-WRT installed). In 2 years it will be legally adult . However I'm aware that most laymen are more afraid of mosquitos that some cyber threats (I'm not implying that anyone in this discussion is a layman).
Both of the devices I'm offering have 1gbit/s ETH ports and the Netis even has an AC wifi. So they are not obsolete by any means. Even though I couldn't test the real throughput because I have only 50/25 mbit/s internet speed.
Couple years ago OpenWRT stopped supporting 4/32 devices which was sad but I found that entirely understandable. 4/32 was barely supportable even 10 years ago. I feel that the decision to stop supporting 8/64 devices came way too soon.
Well there are certainly devices out there which are edge cases. I cannot judge about them because it don't own a mt76 based device.
This is fine as long as you don't need any better performance/hardware or you don't have bad actors nearby your location. DD-WRT is using very old kernels to support Broadcom devices. There is no real development anymore for those devices. It is like sticking to an no longer supported version of OpenWrt for a device (which is OK also).
But you've decided to sort them out for whatever reason. I do not judge the reason. I'm doing the same.
Here is the point: There are X deveolpers bringing up OpenWrt and it's packages within their free time. Those developers are not there for their whole lifetime. People are leaving projects all time for various reasons: family, job, other interests, disputes, frustration, illness/accident, and so on. How life goes. You do the same.
This is always connected with a loss of knowledge.
So it's up to them to decide how many devices they are able to maintain with their available manpower. Just think about a kernel upgrade. You can be sure that each upgrade break things on certain levels. But sometimes it hits like "move all old devices to DSA and make sure it works".
So you need the manpower (ideally the first developer(s) of this device), the device itself on developers desk and/or testers willing to test. The older the device the less likely it is to have it.
In addition to that. Dev's have more insight in what is comming in future regarding package size and memory usage. If they decide 8/64MB is hard to maintain with the resources they have we have to accept it (IMO).
Yes there are a lot of aspects I did not mention. But it is difficult to write things in a way everbody is able to follow. I'm pretty sure that not many understand the sentence about the kernel and DSA. Now as I thought about I came to the conclusion that it's to technical.
So I add the following regarding DSA (new) / swconfig (old):
The main point picking this as an example was that every fkn device had to be touched to get it migrated to kernel 5.4. And I'm sure that this progress is still not finished respectivley some devices failed getting migrated lacking a developer doing it.
Ow, I thought that the main reason to stop supporting 8/64 devices is the amount of Flash/RAM. I didn't know that the main reason is the lack of manpower to keep supporting that many devices.
I don't know if there is a main reason resp. one factor leading to exact this decission. It is likely a mixture of all circumstances. Think of it like a company making decissions about using its resources the best way they can: abilities, workforce, limiting factors without influence (like kernel growth and kernel RAM usage), real costs (dependening on donations), and so on.
It's never possible to support, as in actively test, all 'supported' devices before/ for a release, nor even irregularly at all. OpenWrt developers only have access to a fraction of the supported devices, with many (most?) device support provided by external contributors (some more regularly involved, others more or less drive-by contributions). Keep in mind that well over 800 devices are currently supported by OpenWrt, it's both economically unfeasible to buy all of them, let alone to test all of them in regular intervals. Device specific bug fixing and continuous testing relies on motivated users of said devices to keep tabs on it and the development in the main branch, to actively report issues and patches.
At best there is some coverage of OpenWrt targets (~platforms) among developers, and even that suffers for older or more exotic ones.
Hardware contributions can help in some cases, if there is active developer interest (be it because it's something interesting/ challenging to work with or for an irk to fix long-standing bugs).
)
. However I'm aware that most laymen are more afraid of mosquitos that some cyber threats (I'm not implying that anyone in this discussion is a layman).