Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Use your own WiFi connection test server in Windows (nu42.com)
14 points by nanis on Sept 13, 2021 | hide | past | favorite | 17 comments


While the project is super interesting, this seems like network horror to me. Not only is something "wrong" with the network (guessing dns.msftncsi.com is blackholed), the author is setting DNS servers statically on his clients instead of using DHCP. If I'm right that the domain is simply blocked, I think it counts as yet another horror that the author did not realise that before buying a replacement WiFi card. It pains me that the blog post never explains the root cause of the issue nor whether it was intentional.


There are also plenty of other wifi connection test domains [1][2] around that he didn't have to buy one.

Apple returns "success" 200, opensuse returns 304 (no-content) and may close the browser tab instantly so you have to fiddle in console to see it working.

  await fetch('http://conncheck.opensuse.org',{mode:'no-cors'})
[1] http://conncheck.opensuse.org

[2] http://captive.apple.com/hotspot-detect.html


> There are also plenty of other wifi connection test domains [1][2] around that he didn't have to buy one.

Strictly speaking, I have enough garbage domains that I could have used any one of them, but when I found a cute short name I grabbed it. ... Hosting it myself means my computer doesn't tell someone additional where I am all the time ... Although since the requests are plain http, that doesn't really mean much.

Also note:

    C:\> curl http://captive.apple.com/hotspot-detect.html | xxd
      % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                     Dload  Upload   Total   Spent    Left  Speed
    100    69  100    69    0     0     69      0  0:00:01 --:--:--  0:00:01   489
    00000000: 3c48 544d 4c3e 3c48 4541 443e 3c54 4954  <HTML><HEAD><TIT
    00000010: 4c45 3e53 7563 6365 7373 3c2f 5449 544c  LE>Success</TITL
    00000020: 453e 3c2f 4845 4144 3e3c 424f 4459 3e53  E></HEAD><BODY>S
    00000030: 7563 6365 7373 3c2f 424f 4459 3e3c 2f48  uccess</BODY></H
    00000040: 544d 4c3e 0a                             TML>.
That LF at the end means now I need to figure out how to get that into the registry (maybe it's trivial but it takes an additional ΔT and ΔW.

As for the opensuse one, I am not sure if the check works with empty content. Again, not hard to figure out, but additional ΔT and ΔW.

Plus, these sites can change whatever they do at any time and if I customize to match them, then I am at their beck and call as opposed to controlling both sides of the equation.


The Wi-Fi upgrade story linked in the beginning is also a slightly weird solution. Rather than buy an mPCIe to m.2 adapter and 1st party Intel (or any modern) card he bought a half mini adapter and rebranded half mini Intel card for twice the price.

Not to say there is anything wrong with a stream of consciousness blog about what one has had fun doing doing lately rather a heads up since this is being shared these fixes are just that, not high effort recommendations or whatnot normally posted on HN.


> for twice the price.

Prices vary over time. Upon seeing this, I checked the current price and it is almost twice as much as I paid at the time and the item is listed as out of stock by this seller.

Indeed, there is a 8265 listed at $21[1] and a converter listed at $8[2]. At the time, I did not know such converters existed.

[1]: https://www.amazon.com/Intel-Dual-Band-Wireless-Ac-8265/dp/B...

[2]: https://www.amazon.com/M-2-PCIe-Converter-Bluetooth-2010-201...


Fair point, maybe some differences due to location as well as it shows in stock for me.

FWIW the mpe-AX3000H is a repackaged ax200 not an 8265 (ac only chip). Doesn't change the price just noting it's a different chip than the 7260/mpe-AH3000H https://www.amazon.com/Compatible-AX200NGW-9260NGW-Upgrade-W...


> setting DNS servers statically on his clients instead of using DHCP.

Yes, because these personal laptops end up being used in coffee shops or hotels on occasion. So, whatever I do just on my home router doesn't help.


"On his clients" meaning his eg laptop. Personally my laptop uses its own preconfigured DNS servers, even (especially!) at coffee shops/etc because the coffee shop's DNS isn't to be trusted but more than that, they are frequently extremely slow. (It does take some fussing if there's a captive portal, but that's easy enough to handle.)


> some fussing if there's a captive portal

`example.com` comes in handy.


That is fair, I suppose, but why use a different DNS server than the default for your home network? I still think there's something fundamentally wrong with a DNS configuration that breaks Windows connectivity tests.

I do hope you're using something encrypted. Plain DNS can be redirected and manipulated quite trivially.


One thing I've struggled with is some devices like to use this connection metric to determine when to switch from wifi to cellular, or switch wifi networks, or disable it completely (Android, iOS, Windows.) There have been many times where I'm actively doing maintenance or troubleshooting my network, and the loss of Internet access will completely kick me off my LAN.


At least on iOS what worked for me to stop this from happening is airplane mode + wifi enabled. It probably won’t stop it from jumping to another network though.


This post was really hard to understand. Also, why wasn’t his dns resolving the default test domain ok? Wouldn’t an easier (and more correct) fix have been to address that? That way any other people on his network wouldn’t have a problem, either.


Yeeeh.

Maybe it's a cheap (laborwise) and dirty solution if you don't want to muck about with networking too much. I set up my own subdomain for replacing msftconnecttest because I had similar troubles with Windows 10 showing not connected in task tray, but was actually connected. Maybe that's what he means by 'got flaky'?


> not connected in task tray, but was actually connected.

That has happened to me as well. This was different. The system would go through connect/disconnect cycles and then disable the card etc.

I thought it was a driver issue because every search resulted in people claiming to have solved their problems using older drivers etc, but I had no real luck with that.

The true cause of the problem is still a mystery to me, but pointing the test to my own server gave me control over both ends and has worked fine for weeks now. Yes, it is trivial, but it is a much more preferable outcome than trying to figure out the real problem/solution using search results like this: https://www.google.com/search?q=7260+connection+problems


Wow, looks like Google has managed to get more horrible in the months I haven't been using it.

Try "7260 connection problems" under Duckduckgo and results are much less spammy.


Yes, it's not good.

> That way any other people on his network wouldn’t have a problem, either.

Given that he's setting DNS servers statically on his own clients and has working DNS resolution on the DHCP-provided DNS server(s), that shouldn't really be an issue for other people.




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

Search: