The thread that sparked this rabbit hole:
https://discussions.apple.com/thread/252387604
TLDR: 2022 and we are still stuck with 100Mbit adapters on macOS
The thread that sparked this rabbit hole:
https://discussions.apple.com/thread/252387604
TLDR: 2022 and we are still stuck with 100Mbit adapters on macOS
I've also had good luck with whatever Caldigit uses in their Thunderbolt Mini Docks (not at my desk to check and see what chipset's in mine).
Professionals need all the stable connectivity they can get.
I was elated to hear that Apple is being forced to abandon Lightning on iPhones, and I'm normally not much for government meddling.
In both cases it simply stopped appearing on the bus, but was clearly "running" to some degree - it would get warm to the touch, just as they normally do.
I've also found that the Thunderbolt connector's lack of a "click" engagement to be a serious issue for storage, network, and even display - every Mac I've seen it in use, the connection has been flaky. It really fucking sucks to whip out the thunderbolt adapter and plug into ethernet for "reliability", set up a transfer of a ton of data, and half-way though shift the machine slightly aaaaaaaaand then the ethernet adapter disappears and your transfer is fucked.
Happy to see them drop that idiotic connector for USB-C, as at least that has some sort of physical retention mechanism other than "hopes and dreams" (ie: rely on proper clearance tolerances between different vendors, on surfaces that will wear with use.)
I’ve had MUCH better results with the Belkin Thunderbolt 3 Express Dock HD (INtel i210) than my OWC Thunderbolt 4 Dock (Realtek RTL8153).
I don't have 2.5Gb network equipment but have tested with iperf between and get around like 900 Mbps and no high CPU, unlike noticeable CPU usage with the cheap 1Gbps USB adapters that use ECM drivers
See also https://gist.github.com/MadLittleMods/3005bb13f7e7178e1eaa9f...
A pertinent question, however, is: how do you know what chips an adapter uses, before buying? I'm in utter dread regarding the prospect of buying a usb hub, since just like in the above post, they're black boxes to me which of course turn out to repackage the same Alibaba junk with 10x markup.
The fanboyism and reality distortion field is very strong. I remember when they came out with timer coalescing and were hyping it as a major accomplishment and selling point. Of course, Windows supported timer coalescing for years before MacOS, but that didn’t stop Jobs from convincing a bunch of developers that this was a novel breakthrough.
Or moving/changing the name of a file while it's open.
Or having a slash in a file name.
I still think being able to move/change the name of a file while it's open is impressive
However, MacOS as an operating system really is a mess. Especially the XNU kernel, which is still an unbelievable amalgamation of disagreeing technology. Remember, MacOS is not natively a UNIX-certified machine: all of it's UNIX compatibility comes from a BSD-based compatibility layer that hasn't really been changed since the late-90s. Oh, and the coreutils? Notoriously garbage. MacOS ships with all sorts of outdated, downgraded, vulnerable and otherwise broken shell utilities. pico instead of nano, zsh instead of modern bash... hell, even something as simple as installing git is a 700mb installation with a mandatory reboot.
I'll give MacOS credit where credit is due (Apple had good design philosophies in the 2010s), but the actual operating system (see: functional network of software components) is truly awful, arguably just as bad as Windows if not worse. Just about it's only redeeming qualities are the things that Apple didn't make (like pf and process management. If you forced me to pick something that I found impressive, I'd have to choose Grand Central Dispatch, but even that isn't terribly impressive. It's mostly as if some Apple engineers decided to iterate on the fairly lackluster Linux process management. It would have been a miracle if they managed to make something worse.
Mandatory reboot? I've never experienced that with the Xcode command line tools.
nano is a GNU clone of pico. pico is OG nano.
bash was replaced with zsh because Apple purged their OS of all GPL3 software (including nano).
The command line tools (except GPL ones) do get updated from FreeBSD and there’s nothing “non-native” about how the kernel works.
Maybe what you really mean is you don’t like UNIX and expect it to work like Linux?
When CMU made Mach, it was a BSD OS. When Next made NextStep on top of Mach, it was a BSD OS. When Apple made OSX with NextStep, it was still BSD.
Nearly all of those people have never seen a filename path written out in text, and wouldn't care if they did.
Citation needed
I did this myself the other week. Mac user. Folder was called ‘Lessons/episodes’ from memory. I only noticed it was weird when my Synology didn’t display the character. Renamed to ‘Lessons & episodes’.
I can understand that appeal.
In the last 5 years I don’t think I’ve had a windows laptop sleep or hibernate as I would expect, not randomly wake up in the middle of the night, in my bag, and just sit there fans at full speed…
I’m tired enough of it to try a Mac…
But it's not Microsoft's fault. The Surface Book I replaced it with works perfectly.
You know what Windows and Linux offer that Mac doesn't? A friggin' Hibernate option. Please, just let me have a button to power off my computer while persisting its state to disk. No, I don't want to have to shut down my computer every time I put it in a bag, that's completely ridiculous and an utter waste of my time having to reset all my tmux panes, vim windows, shell histories... what a UX nightmare. They've made a laptop that's a terror to actually take anywhere with you. Even when the "smart" sleep does (sometimes) finally decide to kick in, it invariably costs 20% of the battery.
When people say that Macs have good power management, what they mean is that Safari is optimized for power consumption relative to Chrome and Firefox.
The default is hibernatemode 3 (RAM stays powered on until battery drops below some threshold, so that wake from sleep is very fast).
The version of “hibernate” you want is mode 25: sudo pmset -a hibernate mode 25
If you routinely keep your computer 'hibernating' for days at a time without plugging it in, you’ll save some battery life this way at the expense of a slower wake-up time because RAM is not kept powered.
See https://en.wikipedia.org/wiki/Pmset for more details.
> Safari is optimized for power consumption relative to Chrome and Firefox
This is a very big difference. Chrome has zero respect for battery life.
Also, you can set the power manager's hibernatemode to your liking (Google search "os x hibernate mode") but there usually no reason to adjust the default (sleep for 3-4 hours, then suspend to disk) given how fast storage is in macs these days.
Macs have been famous for decades for having the best sleep/hibernate functions in the industry and when yours didn't work properly maybe you should have investigated? Or at least not be whinging about Apple over it?
I did / have on multiple devices.
I reach inside and it is like a toaster “oh yeah this guy is dead…”
At this point I always globally disable wake timers in Windows. They're mostly used for automatic updates. (For more fine-grained control, look at the various wake timers in Task Scheduler.)
For example something as simple as copy and paste is command-c, everywhere.
In Ubuntu? control-c or control-shift-c. It is pretty annoying being in autopilot and killing the command line program you are in because you reflexively hit control-c.
Also, readline shortcuts work through out, so control-a will send you to the start of a line. Not with Ubuntu.
macos also have it. that is part of application handling. iterm, alacritty, and bunch of other apps can behave the same. Including Xquartz.
If you're going to use Ctrl for shortcuts, you necessarily run into the issue needing a separate shortcut for copy in the terminal, because Ctrl-C has meant "send an interrupt signal to the process" since at least the 60s.
Fortunately, for people sufficiently annoyed by this, most Linux terminals do allow you to change keyboard shortcuts arbitrarily, so you can have unified copy shortcuts if you want. For a variety of reasons, I find this more trouble than it is worth and prefer sticking with the default.
Having a driver for a given Ethernet controller is hardly a good proxy for sophistication…
is that some of us just really don’t like Windows and really like our Macs.
We’re not fanboys. It’s just a preference. Chocolate vs. vanilla, football vs. cricket, boys vs. girls.
I transfer tons of large videos and RAW photos imported from my mirrorless camera over these network adapters to a NAS.
I've been using these for quite a few months.
I'm not stuck on 100Mbps adapters, but maybe you are.
Usermode vs kernelmode seems irrelevant... if anything, I want the benefits of usermode isolation for more things. Monolithic kernels aren't great for security.
https://gist.github.com/MadLittleMods/3005bb13f7e7178e1eaa9f...
You also seem to be digging pretty hard to try and justify your original position, which was very extreme. No, Apple computers are not stuck at 100Mbps. This would be a very big deal, as tons of creative workflows rely on having multi-gigabit network connections. The outcry would be enormous.
I had never tried to use AirPlay 2 from my laptop over ethernet (I can probably count on one hand the number of times that I've used AirPlay from my laptop at all), but I tested, and it is true that the Music app won't connect to AirPlay devices over this Ethernet adapter (but it will over the 8153). I also have no problem redirecting the system sound to an AirPlay device, even over this Ethernet adapter, as that comment says.
It's a completely positive outcome for end users, since they now have more options, and pretty much all 10Gbps ethernet ports are compatible with "multi-gig" 2.5Gbps and 5Gbps connections as well.
10Gbps is effectively the end of the line for copper anyways... 10Gbps is already really hot and power hungry, and AFAIK, datacenters never really bothered to deploy any copper faster than that. Fiber is the present reality in datacenters. What I really want is for SFP+/QSFP+/QSFP28 to make their way into home networks and consumer devices.
This is the basis for my speculation about planned obsolescence.
Nothing presented so far even comes close to making me believe this is a conspiracy. People just got tired of waiting for 10GbE to come down in price. Cost and impatience are the driving factors. Gigabit ethernet was introduced in the late 90s... it was time for some increase in speed, and basically no one has been able to justify the continued high cost of 10GbE in home networks yet, 20+ years after 1GbE, so 2.5GbE it is.
I think the current dilution of 10GBE products by 2.5GBE hardware is a bad thing. I don't want to upgrade my switches every few years when the higher speed interfaces become more common.
10GBE hardware prices have not dropped in the same manner as most other computing hardware because the market is largely "enterprise" customers, and they'll willingly pay more.
Perhaps you are right and most SOHO consumers want 2.5GBE more than 10GBE. I am not one of them.
As an annoying related topic, it's also interesting that although there has been a PoE standard for 10GBE copper for a long time, there doesn't seem to be ANY hardware that supports it. There are however, plenty of 2.5GBE PoE products available. This trend also supports an argument for planned obsolescence, but does not prove anything by itself.
You keep using the term "planned obsolescence", but it doesn't mean what you seem to think it means. Intel producing better processors every year isn't planned obsolescence... it's just the progress of technology, which naturally makes old technologies obsolete. It can't "amount to the same thing" as planned obsolescence. It's either planned or it isn't. 10gig has not reached the consumer space yet. It doesn't matter that it was standardized a long time ago. It's not a conspiracy or planned obsolescence... it's just price vs benefit, and 2.5gig is cheaper because it uses simpler technology.
Calling something "planned obsolescence" is a fairly serious accusation of intent. It obviously annoys me when people make statements like this without evidence.
1) Poor affordability/availability of 10GBE despite its big head start over 2.5GBE. (The 10GBase-T standard was released in 2006, while the 2.5GBase-T standard was released in 2016.)
2) The flurry of 2.5GBE products we are now seeing. The number of 2.5GBE products available now far exceeds the number of 10GBE products available.
3) Zero availability of PoE on the few 10GBE products that do exist despite the generic PoE standard being in place for nearly two decades.
4) The abundance of 2.5GBE PoE products.
Your explanation is "market forces", or in other words, it's cheaper to deploy 2.5GBE than 10GBE because you (may) need to upgrade your cabling when you switch to 10GBE. But is it really cheaper? 10GBE is the end game and we could be there now, but the "market forces" that would normally bring down costs aren't happening because of the artificial scarcity of available products. Cable is the least expensive element of networking hardware, and it may not even need to be upgraded in many cases. (Note that this situation is nearly identical to the transition from 100Base-T to 1000Base-T about 20 years ago.)
Paying for 2.5GBE infrastructure that will be obsolete (or is already obsolete) and then paying again just a few years later for 10GBE infrastructure does not save anybody money. Both SOHO and enterprise consumers will end up paying more. The winners are the network hardware producers.
https://www.eetimes.com/debunking-10gbase-t-myths/ (published 10 years ago)
https://www.microsemi.com/document-portal/doc_view/136209-ne...
I never talked about the price of Ethernet cable, or whether people would need to upgrade it. It’s honestly irrelevant, and a strawman.
10gig is expensive because 10gig is expensive… the switching hardware, the chipsets, the PHYs, everything but the cable. Device manufacturers and customers don’t care about the cost of the cables, especially when 10gig hardware is backwards compatible with lower speeds. They can keep using 1gig if they don’t want to upgrade, just like people continued using 100Mbps networks for quite awhile after gigabit Ethernet became a thing.
My argument is that everyone got tired of waiting for 10gig to be affordable. They waited decades, hoping the price drop was just around the corner so they could jump straight from 1gig to 10gig, and it never happened. After literally 20 years since 10gig was standardized, I’m extremely glad that we’re seeing an abundance of Ethernet hardware that is faster than 1gig. That stagnation had to come to an end.
You can start your own company providing affordable 10gig hardware and prove that the industry players are wrong. For the same price, everyone would snap up 10gig hardware in a heartbeat.
Your argument would have made sense 20 years ago when it seemed like Ethernet standards were rapidly evolving and 10gig adoption was just around the corner. 10gig will be affordable eventually… but 20 years was too long to wait for it, and it still didn’t happen.
If you want 10gig, it has been attainable for years… for the right price. But there’s no need to come in and bother other people who are benefiting from the low cost rollout of better-than-1gig technology.
I would love to have a 10gig or 25gig network, if someone wants to pay for it. Datacenter-class networking hardware (like ConnectX-7) operates at 400+ Gbps today. The sky is the limit, so why bother stopping at 10gig?
> Zero availability of PoE on the few 10GBE products that do exist despite the PoE standard being in place for nearly two decades.
This is incorrect. 10GBASE-T only gained support for PoE in 2018 with the IEEE 802.3bt-2018 standard.
2.5GBASE-T and 5GBASE-T do run over existing Cat5e cabling reliably.
It works fine over Cat6 yeah, but if you've not got that deployed then 2.5Gb or 5Gb is a good step up from 1Gb.
At any rate, a big issue is that for USB, macOS only has generic ECM and NCM drivers. ECM as a protocol sucks and was barely suitable for 100Mbit, let alone gigabit, plus Realtek's implementation of ECM is quirky to say the least.
RTL8156 implements an NCM endpoint, so that's probably the best USB option these days.
That’s something of an exaggeration. There are a bunch of macs that ship with perfectly good 10GigE ethernet adapters - my Mac Pro's ethernet has been rock solid since the day I bought it.
It seems like you’re more specifically concerned with the quality of support for USB<->ethernet adapters, which is always going to be kind of a crapshoot if you’re looking for 24/7/365 connection stability (on a laptop?)