Wi-Fi Issues Continue For OS X Users Despite Updates

itwbennett writes: Although Apple has never officially acknowledged issues surrounding Yosemite and Wi-Fi connectivity, the company is clearly aware of the problem: Leading off the improvements offered in the update 10.10.2 update released Tuesday was 'resolves an issue that might cause Wi-Fi to disconnect,' according to the release notes. Despite this, Apple's support forum was filled with tales of frustrated users. And Mac owners aren't the only Apple users experiencing wireless connection failures after updating their OS. Wi-Fi connectivity issues have also dogged iOS 8 since Apple released the mobile OS on Sept. 17.

Comments

much better than before either way

[Anonymous Coward]

the "broken" state was actually a fix for a much more broken state... unless you want your laptop battery constantly drained while in sleep mode, the wifi antennae is turned off... when you wake up the computer, it used to just assume it was still connected and send out a "hey, i think i'm connected to you" message, which your wifi router would happily acknowledge, and the connection would be instantly available as if the computer never went to sleep. no new logon or auth handshake, just instant access, which is what most users want.... but that opens up rogue wifi hotspots from catching that "hey, i think i'm connected to you" message, and responding with their own "yup, that's me, we're connected" on a slightly more powerful signal which would always beat out the response from the real router, and now you can sniff all of their traffic and steal session cookies. classic airport or coffee shop hack.

so, the "fix" is to require a new auth handshake whenever recovering from sleep mode. that is a good thing. for me, it means i open my laptop and get a "the internet is not available" error if i immediately refresh a web page... but the connection automatically reconnects within about 5 seconds.... that is the kind of thing steve jobs wouldn't have let ship... but the alternative is a massive security hole.

"Wi-Fi" is fundamentally broken, period.

[allquixotic]

The problems with "Wi-Fi" are numerous. The end result is that generally speaking, Wi-Fi is a hot mess of broken tech that doesn't work. In the rare case that it DOES work, even the most trivial of changes in the environment or in the client can completely break it.

1. Early versions of the spec were too loosely worded, and allowed for too many "interpretations".

2. Vendor extensions are still a major problem. Many vendor extensions are not compatible with one another, and a device that has a vendor extension enabled
may work very poorly (or not at all) with a device lacking said extension.

3. Actual implementations of Wi-Fi are all over the map in terms of quality, with ridiculous things like: advertising support for an extension that it doesn't actually support; criminally severe bugs in a production implementation; vendors that try to work around bugs that other vendors introduced but in turn create yet more bugs, causing a vicious cycle of workarounds to workarounds; "hide and go seek" with extensions and spec interpretations; ridiculous driver implementations that hold exclusive access over very coarse-grained locks in the OS kernel for long periods of time, causing freezes and/or panics; poorly designed antennas; buggy firmware that never gets updated; etc.

4. The spectrum WiFi uses is open to be used by literally anything else that complies with a few simple rules, such as the maximum Tx power on that frequency band. As a consequence, random electric devices can freely leak a certain amount of random interference (noise) in the 2.4 GHz and 5 GHz WiFi bands, which destroys the ability for WiFi to operate. Ever lose your WiFi when you turn on your vacuum cleaner, or microwave? That's what's happening.

5. The spectrum WiFi uses is used by other communications protocols that are not Wi-Fi. While some effort is made to interoperate between a few of them, such as cooperation between Bluetooth nodes and WiFi nodes (such that they don't "trample over" one another if they use the same frequency), the interoperation protocols, specifications, and implementations have the same problems as the Wi-Fi specs themselves, as stated above.

6. Recent increased focus on power saving has caused some rather extreme power saving techniques to be employed in Wi-Fi firmware and drivers, which sacrifices performance, range and reliability for a few microwatts or milliwatts of energy. Paradoxically, some of the proponents of these techniques actually think that's OK, and are still trying to make the problem worse.

7. There are a large number of complex physical parameters that affect whether two WiFi transceivers will be able to communicate, which 99% of users don't understand at all. The power saving techniques mentioned above reduce the variety of possible configurations (that is, device orientations and distances, mainly) under which the signal will be reliable and high-performance.

8. Vendors that produce Wi-Fi transceivers, or products that integrate them, usually perform inadequate testing to certify the device as interoperable with a very large array of existing and upcoming other products that use Wi-Fi. Especially in the case of smartphones, the possible number of clients and basestations that may be interacted with is tremendous: Smart TVs; DSL modem/routers; cable modem/routers; other smartphones; enterprise APs and repeaters; laptops; tablets; cars; IoT devices -- all these things need to be tested. With a LOT of work -- and I mean a LOT -- eventually a Wi-Fi stack can be designed in such a way that it operates at least decently well with all modern incarnations of the above, but that says nothing about older implementations, which people love to keep around for a decade or more, and expect them to work. A sufficiently general Wi-Fi stack that works okay with all of the above will probably have so many heuristics for bug detection, compromises, polling tests, etc. that they won't work especially well even in an "ideal" scenario, and may even try to implement contradictory rul

Re:"Wi-Fi" is fundamentally broken, period.

[drinkypoo]

The problems with "Wi-Fi" are numerous. The end result is that generally speaking, Wi-Fi is a hot mess of broken tech that doesn't work.

Oddly, I find Wi-Fi to be the only wireless technology which does work reliably. Bluetooth? It's pure bollocks, it never works. Well, that is, when you're trying to do anything the least bit fiddly. And they permitted all kinds of things which abuse the shit out of the spec to call themselves bluetooth, like PS3 controllers. And my problems with cellular technology are legion. But Wi-Fi tends to just keep on ticking along for me. Routers, those shit themselves, but completely and not just the Wi-Fi.

Actual implementations of Wi-Fi are all over the map in terms of quality, with ridiculous things like: advertising support for an extension that it doesn't actually support; criminally severe bugs in a production implementation; vendors that try to work around bugs that other vendors introduced but in turn create yet more bugs, causing a vicious cycle of workarounds to workarounds; "hide and go seek" with extensions and spec interpretations; ridiculous driver implementations that hold exclusive access over very coarse-grained locks in the OS kernel for long periods of time, causing freezes and/or panics; poorly designed antennas; buggy firmware that never gets updated; etc.

You can have all the same equivalent problems with normal Ethernet, though. Many vendors still sell ordinary 100bTX(etc) chipsets which are total garbage, and people still buy NICs using them and stick them into their machines. I have a whole raft of sketchy tulip clones here because they were $1/ea some years ago at Fry's, back when that was still a really amazing price and not just what you'd expect on eBay. They work OK if you only put one of them in a system, which pretty much crapped on my plans for them, but over the years they've occasionally come in handy. Problem is, PCI is goin' away...

The spectrum WiFi uses is open to be used by literally anything else that complies with a few simple rules, [...] Ever lose your WiFi when you turn on your vacuum cleaner, or microwave? That's what's happening.

Well no. My vacuum and microwave are both brand-name products which seem to have adequate shielding, they do not seem to negatively impact my 2.4 GHz Wi-Fi in the least. In case you're wondering, the vac is a Dyson.

If USB and its "device class" specifications (Mass Storage, Battery charging spec, RNDIS, audio class, etc.) is a ringing success story of how standardization can promote interoperability,

It isn't. Most devices don't use the generic driver, even when it would be better than whatever they actually did.

What sucks about Wi-Fi is that when you have a lot of networks near one another, they crap on one another harder than they are supposed to. It takes extra-special care to make this not happen, and then any dillhole who brings his own battery-powered AP with him can still totally hose your network. But otherwise it mostly works, and often better than the other stuff that's supposed to be better. I had way less radio-related problems when my WISP used Wi-Fi than I do now that they're using some custom CDMA crap. No weather problems either, even though I'm more than five miles from the site. It proves nothing of course, except that the Wi-Fi is more mature technology than whatever they were using.

Now, I have yet to try Chromecast, Miracast etc. but I'm starting to think harder about that, so maybe I will finally learn to hate Wi-Fi. But in general, the only real hardship I've had is with chipsets whose drivers don't support master mode under Linux. This taught me to buy intel and Atheros. Even then you don't get a guarantee, but odds are great and you can always check ahead of time. The thing is, that's good advice anyway, both what to buy and what to do before you buy.

Same on iOS 7

[angel'o'sphere]

My iPad 2, running iOS 7 gets disconnected in certain WIFIs quite often.
My iPhone running iOS 6 stays in the same network without any problems.
Since Mac OS X 10.5/10.6 and iOS 5/6, more precisely since roughly half a year, perhaps a year, before Steve Jobs death, the quality and usefulness of Apple Software goes downhill rapidly ... 10.7 does not even offer a hotkey or a menu to switch desktops. And how to activate multiple desktops, you need to google for. I guess a huge amount of people who never had 10.6 don't even know that 10.7 and later supports multiple (virtual) desktops.
Support for AppleScript, for Apples own programs, like TextEdit, is down to a bare minimum ... you can not even insert a clickable link into a text document via the applications API (not sure if that ever worked, though).
My Application lost the ability to search a few days ago, also the intelligent folders (which I rely heavily on in my business) stopped working. A restart of Mail.app changed nothing. I had to reboot.
Backups to time machine, don't work minimum once a week. Error is: (insert computer name) bundle in use. I wonder by whom? Usually I have to reboot the time machine, which you only can do by unplugging it from the power: which is a bad idea for a device/appliance containing a hard drive. (This is all 10.7 on a MacBook Air)
The back ups for my other laptop, running 10.6 got lost half a year ago. Timemachine gave an error: incremental back up not possible, need to make a full backup. Wow, never had I assumed that implies all my old backups are gone.
I could rant for ever ...
MS I mainly hated because the UI is not useable, and the random bugs that showed up REGULARLY.
Now I start hating Apple, too. At least the UI still works better (for me) than Windows.
Probably it is time to build up my own unix based backup solution and write my own apps for basic stuff like Mail and Web. I can not get it that it is still impossible to freeze a browser tab that is in the background completely ...
A few tabs open in Chrome and Safari and the laptop just runs 3h instead of the 'normal' 10h.

Re:No shit

[AaronW]

Several years ago I worked at Atheros working on improving their Linux driver performance for wireless access points. One of my tasks was to figure out why OS X was quite a bit faster than Linux and Windows. I tracked it down to the way OSX sends TCP ACK packets. Both Linux and Windows send a TCP ACK after every other packet whereas OS X would start to space the ACKs out if the connection was reliable. This had a significant impact with 802.11N where there is a lot of packet aggregation. I discovered that OS X would start spacing the ACKs out to every 16-32 TCP packets instead of every other packet. I wish Linux would do something similar.

I can't comment on any other drivers or current OS X since this was a bit over four years ago.

I can think of a couple of things that would improve Linux for wireless. First of all, spacing out the ACKs like OS X would help. Second of all, supporting transmitting a group of packets at a time, especially to a particular destination, would help a lot. This is due to the way 802.11N aggregates a lot of packets together into a single wireless packet based on the next wireless hop destination. It's much more efficient for cache utilization and for the code paths when groups of packets are handled rather than individual packets since there is a lot of queueing and dequeeuing going on inside the driver.