Sorry, my openwrt information seems to be incorrect and more vendors use openwrt then I realized. So, I really don't know the numbers here.

I do agree with the idea that fixes should be pushed to the mainline and that incremental upgrades should be standard practice.

Arista's SW VP gave a talk where he said that 80% of their customer calls about bugs were already fixed but their customer wasn't following an upgrade policy. This approach applies to most any sw based product.

Bob
Hi David,

The vendors I know don't roll their own os code either. The make their
own release still mostly based from Linux and they aren't tied to the
openwrt release process.

I think GUIs on CPEs are the wrong direction. Consumer network
equipment does best when it's plug and play. Consumers don't have all
the skills needed to manage an in home packet network that includes
wifi.

I recently fixed a home network for my inlaws. It's a combo of
structured wire and WiFi APs. I purchased the latest equipment from
Amazon vs use the ISP provided equipment. I can do this reasonably
well because I'm familiar with the chips inside.

The online tech support started with trepidation as he was concerned
that the home owner, i.e me, wasn't as skilled as the ISP technicians.
He suggested we schedule that but I said we were good to go w/o one.

He asked to speak to my father in law when we were all done. He told
him, "You're lucky to have a son in law that know what he's doing. My
techs aren't as good, and I really liked working with him too."

I say this not to brag, as many on this list could do the equivalent,
but to show that we really need to train lots of technicians on things
like RF and structured wiring. Nobody should be "lucky" to get a
quality in home network.  We're not lucky to have a flush toilet
anymore. This stuff is too important to rely on luck.

Bob
On Oct 11, 2023, at 3:58 PM, David Lang <[email protected]> wrote:

On Wed, 11 Oct 2023, rjmcmahon wrote:

I don't know the numbers but a guess is that a majority of SoCs
with WiFi
radios aren't based on openwrt.

From what I've seen, the majority of APs out there are based on
OpenWRT or one
of the competing open projects, very few roll their own OS from
scratch

I think many on this list use openwrt but
that may not be representative of the actuals. Also, the trend is
less sw in
a CPU forwarding plane and more hw, one day, linux at the CPEs may
not be
needed at all (if we get to remote radio heads - though this is
highly
speculative.)

that is countered by the trend to do more (fancier GUI, media
center, etc) The
vendors all want to differentiate themselves, that's hard to do if
it's baked
into the chips

From my experience, sw is defined by the number & frequency of
commits, and
of timeliness to issues more than a version number or compile
date. So the
size and quality of the software staff can be informative.

I'm more interested in mfg node process then the mfg location &
date as the
node process gives an idea if the design is keeping up or not.
Chips designed
in 2012 are woefully behind and consume too much energy and
generate too much
heat. I think Intel provides this information on all its chips as
an example.

I'm far less concerned about the chips than the software. Security
holes are far
more likely in the software than the chips. The chips may limit the
max
performance of the devices, but the focus of this is on the
security, not the
throughput or the power efficiency (I don't mind that extra info,
but what makes
some device unsafe to use isn't the age of the chips, but the age of
the
software)

David Lang

Bob
On Wed, 11 Oct 2023, David Bray, PhD via Nnagain wrote:

There's also the concern about how do startups roll-out such a
label for
their tech in the early iteration phase? How do they afford to do
the
extra
work for the label vs. a big company (does this become a regulatory
moat?)

And let's say we have these labels. Will only consumers with the
money to
purchase the more expensive equipment that has more privacy and
security
features buy that one - leaving those who cannot afford privacy and
security bad alternatives?

As far as security goes, I would argue that the easy answer is to
ship
a current version of openwrt instead of a forked, ancient version,
and
get their changes submitted upstream (or at least maintained against
upstream). It's a different paradigm than they are used to, and
right
now the suppliers tend to also work with ancient versions of
openwrt,
but in all the companies that I have worked at, it's proven to be
less
ongoing work (and far less risk) to keep up with current versions
than
it is to stick with old versions and then do periodic 'big jump'
upgrades.

it's like car maintinance, it seems easier to ignore your tires,
brakes, and oil changes, but the minimal cost of maintaining those
systems pays off in a big way over time

David Lang

-------------------------

Nnagain mailing list
[email protected]
https://lists.bufferbloat.net/listinfo/nnagain

-------------------------

Nnagain mailing list
[email protected]
https://lists.bufferbloat.net/listinfo/nnagain
_______________________________________________
Nnagain mailing list
[email protected]
https://lists.bufferbloat.net/listinfo/nnagain
_______________________________________________
Nnagain mailing list
[email protected]
https://lists.bufferbloat.net/listinfo/nnagain

Reply via email to