That used to require a crossover cable; I've done precisely that (with a crossover cable) back before https://en.wikipedia.org/wiki/Medium-dependent_interface#Aut... became widespread. With a straight-through cable, you'd be connecting the Transmit (TX) pin of one adapter to the TX pin of the other one, and the Receive (RX) pin of one to the RX pin of the other — and neither device would "hear" the messages the other one was sending. A crossover cable flipped those wires, so each device's TX pin was connected to the RX pin on the other end, and both devices could "hear" each other.
But with auto MDI-X, each device would notice "hey, I'm sending but not receiving anything," and would try flipping its Transmit and Receive functions around (transmitting on the RX pin and receiving on the TX pin). Since each device waited a random period before doing that, it was very unlikely (nigh-impossible) that they would both flip at the exact same moment. And if they did, the second interval would most likely not be identical either.
I'm simplifying a bit in the explanation above, but that's the broad strokes. And that's how my carefully-labeled crossover cables started gathering dust. (And then I realized "hey wait, I can just use these as normal cables now", and pulled them back out of storage and mixed them with my normal patch cables).
Last time I used a crossover cable I was still using dialup, probably about 20 years ago. Once I had DSL I also had a router, so the crossover cable became redundant.
I believe a lot of NIC's could autodetect and adapt to the cable? I'm pretty sure I had 2 computers connected to each other using a regular network cable.
Interesting. I don't remember my crossover cables being marked that way; you had to read the dot-matrix printing along the side to know. Color would have been way more handy.
I added labels (just a piece of paper under some clear tape) saying "CROSSOVER" at both ends of the cables, so that if I ever plugged a crossover cable into a place where I needed a straight-through, the label would be right there next to the plug.
Well I have to add that was interesting too. Because I remember you had to hold the ends up next to each other and compare the order of the colored wires. Same order left-to-right, not cross-over. Anything else, cross-over.
Oh, I had no idea you could perform such a "broadcast" ping on link-local addresses. Now, that is a nice trick that will save me some typing between computers.
mDNS / Avahi is also great if the machine advertises itself: just use machinename.local after plugging it.
Though the interface needs to be configured for link-local addresses. In my experience, NetworkManager now periodically resets the link if it can't get a DHCP server to respond. I understand the rationale, but it used to be simpler to plug into a headless machine and be confident that you could just ssh in.
I would assume so as well, I'm not sure if all host operating systems act the same though. But you could probably just wait a couple of seconds and the interfaces would have assigned themselves link-local addresses.
> I'm not sure if all host operating systems act the same though
I have two USB-C Ethernet adapters laying around that I bought some time ago. I connected one of them to a Linux machine and the other to my MacBook Pro and experimented a little bit.
All of the following commands and outputs are from the MacBook Pro. The USB-C Ethernet interface is en8.
On the macOS side, I can see that a 169.254.xx.xxx/16 IPv4 address has been assigned by macOS as there is no DHCP running on the link, and it has an fe80::xxxx:xxxx:xxxx:xxxx%en8/64 IPv6 address assigned by macOS as well. (The x-es in the aforementioned IPv4 and IPv6 addresses were put there by me in this text. On the system they are decimal and hexadecimal digits in the IPv4 and IPv6 addresses respectively.)
The first attempts, relating to ssh, maybe would not work on any system anyway? Not sure how the ff02 multicast prefix for link-local scope works exactly and what its limitations are.
% ssh ff02::1%en8
ssh: connect to host ff02::1%en8 port 22: Address family not supported by protocol family
% ssh -6 ff02::1%en8
ssh: connect to host ff02::1%en8 port 22: Address family not supported by protocol family
% ssh -B en8 -6 ff02::1%en8
ssh: connect to host ff02::1%en8 port 22: Address family not supported by protocol family
% ssh -B en8 -6 ff02::1
ssh: connect to host ff02::1 port 22: Address family not supported by protocol family
I then tried a couple more variations with ping similar to the variations I did with ssh, before moving on to ping6. Seems that the ping command on macOS is IPv4 only. I don't see any -6 flag for the ping command in the man page on macOS. But there is the ping6 command for IPv6, which I have used before (just not with ff02 addresses).
Finally, ping6:
% ping6 ff02::1%en8
PING6(56=40+8+8 bytes) fe80::xxxx:xxxx:xxxx:xxxx%en8 --> ff02::1%en8
16 bytes from fe80::xxxx:xxxx:xxxx:xxxx%en8, icmp_seq=0 hlim=64 time=10.893 ms
16 bytes from fe80::yyy:yyy:yyyy:yyyy%en8, icmp_seq=0 hlim=64 time=14.365 ms
16 bytes from fe80::xxxx:xxxx:xxxx:xxxx%en8, icmp_seq=1 hlim=64 time=0.511 ms
16 bytes from fe80::yyy:yyy:yyyy:yyyy%en8, icmp_seq=1 hlim=64 time=2.467 ms
16 bytes from fe80::xxxx:xxxx:xxxx:xxxx%en8, icmp_seq=2 hlim=64 time=0.440 ms
16 bytes from fe80::yyy:yyy:yyyy:yyyy%en8, icmp_seq=2 hlim=64 time=2.393 ms
^C
--- ff02::1%en8 ping6 statistics ---
3 packets transmitted, 3 packets received, +3 duplicates, 0.0% packet loss
round-trip min/avg/max/std-dev = 0.440/5.178/14.365/5.422 ms
Worthy of note is that when I ping6 ff02::1%en8, I get responses from both side of the connection (the ones I substituted with fe80::xxxx:xxxx:xxxx:xxxx%en8 are the same IPv6 address that I see for the en8 interface in ifconfig output on my MacBook Pro, and the ones I substituted with fe80::yyy:yyy:yyyy:yyyy%en8 are the IPv6 address of the other end of the link-local connection).
Attempting to ssh to the IPv6 address from the ping6 output, making sure to pick the one that is not the IPv6 address of the ethernet interface that is connected on the MacBook Pro side itself:
ssh fe80::yyy:yyy:yyyy:yyyy%en8
Output of that I will omit for brevity, but it is the standard question about accepting the ssh fingerprint of the machine and letting me know that this fingerprint is previously known by another name, as I have previously ssh'd into that machine from this one but using the .local mDNS hostname rather than this fe80 IPv6 address.
So it seems that having ping6 ff02::1%en8 up the sleeve is useful indeed on macOS too, as long as one knows to use ping6 and not just ping, and that attempting to ssh to ff02::1%en8 does not work directly from macOS at least not for me.
"It's possible to just connect two computers together with Ethernet" - on the front page of hackernews. Incredible. Ethernet over USB4 interdomain protocol at least had some novelty, but this...
Programmers have very little understanding of computers, in general.
They'll think hard about a beautiful unrealistic idealisation of the problem, and/or treat a software abstraction as if it's the underlying reality, e.g. (of the latter) thinking networking starts with HTTP and languages start with JavaScript.
It does seem to be a surprisingly common blindspot of programmers in general. I strongly recommend programmers learn it, it can save a lot of headaches in system architecture.
Back in the day you needed a modified cable for this with RX/TX swapped on one end but nowadays probably all nic chips have auto-detect so you can take a normal patch cable and do it.
I bought bunch of ready made 10 meter cables for lan party dunno 15 or 20 years ago... Used with switch... Grey with nice paper label at end with "Cross over"...
Haven't had any issues with them ever in regular use... So pointless even back then...
When I hacked a USB2-GigE adapter into my old Toughbook CF-m34 to replace the stock 10/100 adapter, it wasn't for the speed. I didn't need the speed, and of course USB2 wouldn't let me use the whole gig anyway.
But GigE uses all four pairs, so the hardware has transceivers behind all four pairs, which means even when they're running at 10 or 100, the chipset can choose which pairs to transmit and receive on. Thus all GigE adapters are Auto-MDIX, and it meant I no longer needed to carry a crossover cable, no matter what other device I was talking to.
It's funny how 10/100 cards with Auto-MDIX capability were exotic and expensive, but once GigE had it by default, it became ubiquitous, cheap, and quickly forgotten.
Exactly. No need to get cynical about this being news. Maybe not news to you (or me, I'm old) but it could be really interesting news to others. See also: https://xkcd.com/1053/
On related news, you can talk point to point with just a pair of copper wires that physically connects with a device on the other side!
Wait there's more: a couple of simple plastic glasses joined with a string under tension also does the trick.
However more seriously, while being snarky about "obvious" things can be fun, we should all remember to acknowledge the xkcd's "Ten Thousand" effect as a very real law of life.
This remind myself the good old 8bit days of using a cross-over RS232 cable to send a file from one computer to another. even at 30bps, it was much more reliable than write my data to a cassette tape on one computer and then reading on the other.
232 crossover was still going in the PC era with laplink. Parallel was an option too if I remember.
Used to play Duke Nukem 3D over a null modem serial cable.
In the days of the spectrum where you loaded via an analog audio input from a tape, and saved via the output, you could move from one to another in the same way, skipping the tape.
Because if you had modems between the computers, you used straight-through RS232 cables -- the PC would have its Tx/Rx pins according to DTE standard, and the modem would have its Tx/Rx pins according to DCE standard. The modems did the "swap" themselves by choosing tones at dial/answer/negotiate time, then the other modem-to-terminal link was likewise straight-through.
But if you were connecting two terminals directly with no modems, thus the modem was "null", you needed the cable to do the signal swap instead. Hence the name!
I have plugged two ZX Spectrums together and done LOAD "" on one and SAVE "THING" on the other to transfer data.
If you get away from BASIC commands and copy the tape routines up into RAM you can mess with the timing values, and with care you can get it from an average of 1800bps up to around IIRC 12000bps before things really fall apart. Probably with better interfacing you'd get even faster.
With more conservative values, this is how "speedloaders" worked. In the late 80s or early 90s one publisher released games on CD where it would load in the loader at normal rate then ramp up to some unholy speed that CD had the bandwidth and stability for but tape did not.
You can also pipe it through zstd on the fly, for data that compresses well you'll often see 1.5 to 3× the raw throughput, so a gigabit link can effectively move 165–330MB/s.
-T0 uses all cores. Bump the level above -6 for more compression, drop it for more speed, but if your CPU can't keep up, high levels will actually slow it down. Already compressed data won't see much benefit.
If you're using specifically zstd, on the sender side, instead of tweaking the whole tunnel once, you can use --adapt to dynamically adjust to i/o conditions.
I wonder how we managed to invent the Internet before this trick. Next week's front page will be "Amazing trick: connect two wheels with a platform and you can ride instead of walk"
I literally had to cross over the cable once. It was some sort of demo and the cross over cable was missing. The local computer stores didn’t have one and under the time pressure … I have used the pocket knife and convert the cable. It was before 2000 so the speed was limited to 10/100 Mbit - ifconfig did’t report errors and demo went smooth.
"Plug two computers into each other with a network cable and staticly configure the IP addresses" should not be a novel or notable enough idea to trend on HN! This is boggling.
No ned to bring it up as it should already be up. enX should be your USB/thunderbolt ethernet adapter, e.g. en6.
Then you can go ahead and transfer with socat to your other machine, even if it's a Linux machine. I tend to use nc but socat will definitely be faster.
Why the deadbeef network one might wonder? It's private space so won't conflict with public addresses.
That being said though, in 99% cases of big files I'll just airdrop the files!
ifconfig still probably works on Linux too but it's discouraged because, frankly, it sucks compared to ipconfig2. Most of macOS' CLI feels old and crusty. Which makes sense when you realize it's a fossilized niche BSD from the early 00's.
tbstream looks fun, but hard to use for the casual user -- not sure if it'll get simpler but I don't think I'll remember messing in configfs just to transfer a file:
It would be nice if this article defined a "ethernet patch cable". I think he's just using "patch" as a slang term for a short cable and it actually works with any length of standard cable, but I'm not certain.
Traditionally (pre-2000?), one had to use a special "crossover cable" to do direct connections like this, but apparently modern Gigabit ethernet adapters are able to detect this situation automatically?
https://en.wikipedia.org/wiki/Medium-dependent_interface#Aut... is what allowed the switchover to happen automatically, and it got folded into the 1000BASE-T standard. So yes, crossover cables are no longer required because the two Ethernet adapters at either end can negotiate between them to have one of them "flip" the meaning of its Receive (RX) and Transmit (TX) pins.
They are all just twisted pair cables, probably Cat5e or Cat6 etc. The solid ones are usually used for structured cabling, inside walls and trunking, that doesn't often move. Structured cabling is terminated into keystones or patch panels, the "female" end.
Patch cables are outside of walls and designed to be moved around and plugged and replugged. They are terminated with 8p8c "male" connectors, commonly called rj45. Stranded cable is used because it's much more flexible.
There doesn't seem to be agreement as to exactly what one is, with several definitions already appearing in the comments here. Do you agree with one of them and think everyone else is wrong? Or do you have yet another standard definition that we should all be using?
My problem with the article was that "patch" doesn't seem to add any useful information. Unless he was using a restrictive definition intending to exclude things like crossover cables (which still work), he could have just as well said "transfer files over a blue Ethernet cable" with equal precision.
If he meant to say that at this point in time "you can use _any_ ethernet cable to transfer files directly between two computers", I think it would have been better to say that rather than appearing to add a restriction on the type of cable required.
No, traditionally a patch cable is a cable that you would use in a networking rack to connect (to "patch") from say a network switch to the desk outlets or to another switch. A patch cable would explicitly be NOT a crossover.
The main difference between patch and "normal" is that the former is using stranded conductors and the latter solid conductors.
That makes the patch cable more flexible (physically)
Is the cable connected from say desktop computer to wall socket a patch cable? Or does that only apply to those used in server rooms or switch cabinets?
I have a feeling this is some re-invention of the definitions happening.
It's just a holdover from the telephone networks, where you used a short cable to actually phsyically connect - patch[0] - two lines. Therefore - patch-panel and patch-cord or patch-cable.
Patch cables are differentiated from bulk/structured cabling which is installed into a building unterminated, and is most commonly terminated after installation into patch panels or keystone jacks.
I am not sure that I know of any Ethernet (IEEE 802.3) standard cabling which uses solid-core wires rather than twisted-pair stranded wires. I can understand and see that many people consider "structured cabling" to consist of the former, but this article does not seem to mandate that construction, and indeed includes many types of twisted-pair and "patch" cables in the list of acceptable "structured cabling solutions."
Yeah, it's not patch vs structured. Patch cables are part of structured cabling. But you won't generally find solid core wires outside of structured cabling, because they are inflexible, unwieldly and in many cases can't even be terminated with RJ45 plugs anyway (they are often too thick).
To regular people it's probably enough just to say Ethernet cable. Outside of structured cabling you don't really need to care and probably don't have a choice anyway. If someone says patch cable they probably know a bit about structured cabling and couldn't help themselves.
Within structured cabling the rabbit hole is quite deep. Even within a single standard like Cat6 there are many possible cable constructions with different types of shielding and wire gauge etc. It becomes important when you need the network to certify as Cat6 and support PoE in a high noise industrial environment, for example. And, of course, Ethernet can run over other cable types anyway, like fibre.
A patch cable is just a terminated cable with multi strand leads (as opposed to solid leads in installation cable), it could be straight patch or crossover.
Actually, switch to switch would be a crossover cable (pre auto MDI-X, of course). The rule was a crossover would connect devices of the same type, e.g. computer to computer, hub to hub, switch to switch; and patch for devices of different type, like switch to computer. So if you wanted to just plug a hub into the wall you would use a crossover cable, for example.
The USB/Thunderbolt one is a nice catch and good to know if you want to move you data to a new PC. Dell XPS (with soldered SSD) to Framework move just took a few minutes by dd'ing through the created network interfaces (it's different from the tbstream). The limiting factor was the Dell's SSD actually.
I've been doing this for years between MacBooks with a thunderbolt 4 cable. It's very fast and unlike 10Gbps ethernet, doesn't require a bulky, power-hungry[1] adapter. MacOS also will use thunderbolt automatically for Migration Assistant if available.
[1]Most people don't realise how much power ethernet consumes. Even 100Mbps ethernet consumes about 0.5W per port and it goes up with higher speeds.
Before the world standardized on Ethernet/IP, Microsoft had an abstraction layer called DirectPlay. ("Direct" was that era's "Copilot" word for Microsoft)
FYI, EtherNet/IP is an industrial control protocol, not TCP/IP over Ethernet. Yes, that is the real name of the protocol, a more confusing name could not have been chosen...
When I did LAN parties we had 10base2 over thin coax. Terminators at each end. Of course it was unstable because people kept disconnecting and everyone lost connection.
For anyone today who's got copper wire stretched between any buildings, you are in danger and your configuration is quite unwise.
For every standalone building that has a ground, that building also has an electrical potential. This potential is not necessarily the same between two neighboring buildings. When you stretch a copper line between them, you are tempting fate, electrically, not to mention the dangers of lightning strikes and the elements attacking that connection.
It is generally recognized that fiber optics are not only very efficient and durable, but also impervious to electrical troubles of the type that may plague inter-building connections. Simply grab switches or routers that support at least one fiber-optic connection, and link up!
I did work for a small community college back in the 90s that had expanded their building multiple times since the 70s. The electrical potential between different parts of the building was significant enough to make wired Ethernet unreliable. (I never did understand why this was-- multiple electric service entrances with separate grounds, I assume.)
Because of this they were an early adopter of fiber-based networking. I remember being wowed by ATM over fiber running between closets at a blazing 155Mbps (with tons of 10BASE-T clients connected to FORE chassis switches).
Ethernet is electrically isolated at every port (unless you really go out of your way in to design it otherwise in an internal design). Your advice is still correct but it's more about lightning or something gone wrong, it's not an immediate risk.
I don't remember how it happened, but I have seen faulty Ethernet devices suffering a power surge that burned the switch port they were connected to, so be careful.
Besides you worry too much. Entire cities were wired with cat5 hanged off steel cable between 9 to 15 story buildings in early 2000s, before fiber. Nothing burned down.
If your cable is shielded it's liable to spark and trip breakers. Happened to me when one building was horribly miswired (someone swapped live and ground somewhere) and fried part of my computer before the breaker tripped. In the worst case there isn't actually any breaker in the path (they are only on live wires and the worry here is ground loops) so your cable just melts. That didn't happen to me.
Unshielded cable is theoretically isolated at both ends but who actually knows with some chinesium equipment?
For the time spent finding the ethernet adapter, plugging the cable, configuring IP addresses, and testing with pings, any decent portable SSD drives should have finished copying tens of GB of data.
It seems a strange choice, for the demo and for the script, to manually configure addresses on both ends. If your TCP/IP stack is functioning properly, this will not be necessary. Once DHCP fails, you should get a pair of 169.254.0.0/16 (APIPA) addresses, and then Bob's your uncle.
Configuring all this manually adds extra complexity when it seems that the goal is simply to connect up your cable and let 'er fly.
Back in my college days, 169.254.x.y addresses were the bane of my existence, because my job was getting students' computers set up with the brand-new Ethernet connection in their dorm room. This usually required a tech to come out to their room and configure Windows 95 properly, with IP and DHCP instead of Novell Netware and IPX (I don't remember why Netware/IPX was the default on so many Windows 95 installs, but that was definitely what I found most often when I went to look at someone's computer that "wasn't connecting to the Internet".)
Occasionally we had someone who knew what they were doing and had their Windows 95 settings set up correctly, and then all we had to do was activate their room's Ethernet port and add their MAC address to the DHCP server's list of authorized MACs. But most often I or one of the other techs had to head down to the person's dorm room and set up their computer's network settings correctly before it would work. Once the 169.254 address was replaced by 192.168, my job was finally done. But there were times when that took some doing.
And if you prefer ipv6 you can as well just ping -6 ff02::1%eth0 (or eno1 or whatever your devices name in the particular subnet is) and receive all fe80 addresses of other hosts in the subnet.
Also, why is he talking about "ethernet"? Its the IP layer, not the ethernet layer...
Depends on the device, in principle if both have thunderbolt and you use a thunderbolt cable it should work. I've done it between 2 macbooks many times.
With USB-C this is a lot more likely, the main issue is going to be software support (Decent chance you could hack something up if one of the devices is running linux. No idea if anything exists for windows).
You would need a USB hub that could turn a USB device into two USB device ports, which I don't think exists. (hubs are essentially only a host-side thing).
What a clever idea. I carry an ethernet dongle but this one is just the male end and a female usb-c port, meaning your existing usb-c cable is useful and you don't need to carry an ethernet cable.
Oh. My. God. I thought you were just going to link a network dongle. "Yeah, I have a drawer of those". It never occurred to me that you could point it in the other direction.
I'm ordering a dozen. Okay, maybe 3. But still, those look amazing.
I have been using USB-C dongles like this for a hot minute now paired with a nice retractable travel cable. I have a couple of these, an HDMI male one, a DisplayPort male one - I can plug basically into any hotel TV or monitor in my travels and run a laptop, phone, whatever off of the cables in my bag for charging, data, or video.
But with auto MDI-X, each device would notice "hey, I'm sending but not receiving anything," and would try flipping its Transmit and Receive functions around (transmitting on the RX pin and receiving on the TX pin). Since each device waited a random period before doing that, it was very unlikely (nigh-impossible) that they would both flip at the exact same moment. And if they did, the second interval would most likely not be identical either.
I'm simplifying a bit in the explanation above, but that's the broad strokes. And that's how my carefully-labeled crossover cables started gathering dust. (And then I realized "hey wait, I can just use these as normal cables now", and pulled them back out of storage and mixed them with my normal patch cables).
reply