Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Honestly this reads as painfully naive. Maybe you’ve never worked with radio communications before in which case I definitely give you the benefit of the doubt for not knowing this:

But wifi (just like other radio tech like 4G) has limited capacity, and that capacity is shared across all radio systems- this sounds obvious but when you realise how few devices can be talking simultaneously you will understand what an absolute pain this is.

Most of the huge gains in wifi technology (and 3g-4g) is not just making assumptions about what a signal is based on the leading and trailing edge of the wave of the transmission: mostly it’s about collision detection algorithms.

Collision detection, by necessity, adds undeterminable latency to your connection, it’s literally how it works.

So taking calls over wifi in a room full of people using wifi: you’re going to have a bad time. Worse: that room full of people is not _just_ that room full of people. Wifi works in 3 dimensions, you’ll be competing with your upstairs and downstairs neighbours too.



> So taking calls over wifi in a room full of people using wifi: you’re going to have a bad time

You aren't wrong [1], but at the same time you are missing that it is entirely possible to provision a wifi setup that can support a large number of devices in a small space. In an office setting, you should be incredulous if they don't have it right. But at home, with consumer networking gear (like the WSJ is talking about)? No way, ethernet is going to outperform it - it will be like night and day.

[1] https://en.wikipedia.org/wiki/Collision_domain


Thanks for clarifying the point, I'm sorry I didn't make it clearer.

WiFi devices can support many devices in small spaces, but you will have increased latency and diminishing returns on throughput for the reasons I mentioned.

Downvolting radios will only get you so far.


I can remember when most wired consumer gear were hubs, not switches. The network would become nearly unusable when you had even a tiny LAN gaming party. Now, I'm not sure that you can even buy an ethernet hub.


And yet countless massive office buildings across NYC and every city seem to allow dozens of people to work from a conference room without issue simultaneously over the air every day... must be magic.


>without issue

>...must be magic.

i'd believe the second line if the first one were true, but fact is that although your experience may be very good, people have issues with WiFi constantly -- and more importantly; a lot of people are having issues without even realizing it.

An AP that hits 40% of it's projected performance metrics isn't down, but to say it is issue-free is malarkey.

Air-congestion is very real, and it affects all of us. Average latency is more chaotic now than it ever has been -- even if the low numbers are very low.

Is WiFi effective? Absolutely. I have no illusions that an office worker filling out spread-sheets is going to be totally hindered by a bit less predictable latency numbers -- I just think that it paints an incomplete picture by not realizing that we ARE losing something for the convenience of network-over-the-air.

And just to throw an anecdote out there , i'm often one of those conference room workers in one of the massive office buildings when contracted , and with a brand new XPS I still see all the same old issues with congested WiFi nearly everywhere I go. "Convention Wifi" is such a meme that most people expect it to fail 1st or 2nd day of any specific Con. It's not all that uncommon for me to encounter corporate groups that have specific APs for video conference that aren't allowed for general-employee use; the fact that this trend sprung up in the first place presumably indicates that not everyone is getting the performance they need from the arrangement.


I worked in a purely wifi-based office of hundreds of people. Seamless roaming, speeds in hundreds of Mb/s, low ping - while everyone restored their NPM deps, Docker image layers and had video conferences all the time. Don't know what you're talking about.

IME wifi problems are purely administrator capability issue. And I've never seen anyone have a problem with mobile network due to wifi...


I don’t want to sound flippant but honestly where you worked has somehow defied the laws of physics.

I would recommend spending 50 minutes of your life watching this: https://youtu.be/HRhZniqyey8

While many access points will state that they support 250+ clients, what they mean is that 250+ clients can theoretically be “connected” as in pinging the base station. Actual throughput will be limited.

Having worked in high density setups I simply cannot believe your experience. Even with low power and many APs there would be too much crosstalk/overlap. It cannot be the way you claim it to be unless you were 10 people and had isolation from other APs.


They just had like 100 APs and well configured roaming. I guess they carefully managed power levels and channels of each AP.

It's not much different from 4G mobile data working well at musical festivals with thousands of visitors.

IME from where I live, mobile data work well even when standing in the middle of protests of 10k people (many of them live streaming on FB and Instagram, sending selfies and photos and videos...). 100k becomes a problem though.


I'd love to see this done right, The only company I've seen go close was using ubiquity hardware and they had custom engineering work done to make it work well.

I'd -read- the shit out of the book if someone wrote it.


Same here. These corporate network engineers are wizards.

When I was a kid, my father had a small WISP. Even that (done with consumer-level hardware) was super interesting.

IMHO networking is surprisingly accessible though - it's well documented because the field is built on certifications.


> It's not much different from 4G mobile data working well at musical festivals with thousands of visitors.

I promise sincerely that this is the kind of naivity that I was referring to.

A single cell site has something like 36 "TRXs" (this is independent of 3/4g), but with 4G due to the fact there is better collision avoidance and a lot of work on the trailing and leading edges inferring meaning: you can have 36 calls without increased latency.

You will notice this.


Naivity doesn't matter when you're there and it works. Repeatedly, reliably.


It doesn’t, but it matters when you extrapolate and generalize that Wi-Fi is more than fine for most if not all the people ( even if that wasn’t you intention )


A good RF engineer who specializes in WiFi deployment can absolutely design a solid environment. For the best deployments they're involved during the building design process to identify the general area where access points will be needed based on the expected building occupancy (and just as important, how people will be distributed in the building). Near the end of construction they'll be on-site to take reading and fine-tune the placement. After construction ends they'll often be on-site several times to take additional readings and adjust power levels (although large companies may have people who can take the readings so the engineer doesn't have to come on site to make updates).

And honestly, even most conference facilities are able to provide acceptable WiFi for attendees. --I've participated in meetings where 100 people would be in one room and we all had decent connections.

WiFi has it's problems in environments where the access point owners won't cooperate (home environments), but it's much better than it used to be.


> I've never seen anyone have a problem with mobile network due to wifi

that's either due to ignorance or negligence




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: