why haven't ethernet connectors changed?

I guess a magsafe ethernet connector would have too much noise (owing to
poor quality connection) to provide decently high bandwidth.

This thread reminds me of http://fanf.livejournal.com/96172.html


You didn't include RJ11 in your question.... it goes back further.

One reason is that as we push the limits of cable from CAT3 (10meg) to CAT5
(100meg) to 5E (gig) to 6 (not sure what that was for) to 7 (10gig), the
cable doesn't get any smaller. We're dealing with higher and higher
frequencies of changes on the wire. This makes cross talk and interference
a bigger problem, so the twists and insulation are more important to try to
protect from those issues (sometimes shielding). So the cable hasn't gotten
any smaller. The connector works well enough and allows for these distances
to be maintained. Some vendors have found ways to maintain the twists
farther into the RJ45 by essentially using traces and not just lining the 8
wires up in parallel but stacking them in a staggered fashion...

Obviously, a new connector could have been found, but why haven't we
changed the C13 that HP came up with (at least I think they did) back in
the 50s? Its still the defacto standard for all computer input power. As a
matter of fact, most NEMA specs haven't changed since they were created...

If it ain't broke, don't fix it. The only problem with the RJ45 is the hook.



I'd turn this back the other way though: in this day and age, why do we have any
interconnection/bus that isn't just ethernet/IP? IP, as we all know, doesn't imply
global reachability. What we far too often do with specialized IO channels is recreate
networking, usually poorly.

That too would solve the Raspberry Pi problem.

Mike, naming being one big issue which is getting short-shrift in homenet

Don't bet on fiber to the desktop making any inroads before Amp's
patents on their Lightcrimp Plus system expire. They're the only ones
to get close to making field termination of fiber a casual task with a
low barrier to entry and they're dead set on making the Iomega

Bill Herrin

I don't see why a magsafe connection would be any more or less noisy
than an rj45. They both follow the same principle: spring tension to
hold the contacts together.

The main issues with magsafe are:

1. You can't have very many pins before the power of the magnet
necessary to overcome the spring tension approaches the ridiculous.
2. Past about two magsafe connections to a machine, cable tangle will
cause them to frequently pull loose.
3. RJ45 implements spring tension the simple and cheap way. Magsafe
does it the complicated and expensive way. You can pretty much forget
about field termination.

Want some entertainment? Read this article on repairing a magsafe connector:


Bill Herrin

That's what cable boots are for.

Bill Herrin

Please, no connectors that do not lock into place. Is plugging in the
RJ-45 that much of a task? Most portable devices are going wireless in
any case so they are not an issue. The RJ-45 has worked OK for me. The
AUI connectors have a special place in networking hell. What an
incredibly horrible mechanical design they were? The flip side of the
question is why you think the RJ-45 should change. You could argue that
you don't usually need all eight wires but every time we tried that
argument someone came up with a compelling reason to use more wires. I
like that it is very standard. In the fiber world it is a continuous
issue of hybrid patch cords dealing with ST,SC,LC and all the other
variants out there. It would be a huge nightmare if the same thing
happened with copper Ethernet.

I am also not a huge fan of the USB connector because I have seen a lot
of those break and there is no positive retention. Magnetic is cute but
has no place in a datacenter and even with desktops I can picture a lot
of support calls because someone bumps a wire that knocks the mag
connector out of place. I really hate dongles of all types but I guess
you don't really have a choice with devices so physically thin that you
can't get the jack in there.

I think I will keep the RJ for now.

Steven Naslund

The need for isochronous transmission and more bandwidth.


The only thing I would change about RJ-45 is a longer tab (but make it optional) for when you care more about ease of removal than cable tangles. Polycom phones are hell to try and unplug the RJ-45, for example.

That's why G*d invented RTP, of course. And all of these buses are "slow"
by the time they're popular enough to worry about. In any case, delete
the "ethernet" part if you want to still play with the mac/phy.


Once upon a time, Michael Thomas <mike@mtcc.com> said:

That's why G*d invented RTP, of course. And all of these buses are "slow"
by the time they're popular enough to worry about. In any case, delete
the "ethernet" part if you want to still play with the mac/phy.

Well, the reply was sent in response to somebody talking about HDMI.
HDMI 1.4 can carry over 8 gigabits per second, so to re-use ethernet PHY
(and still be copper) you'd have to go with 10GBaseT. The cheapest
10GBaseT card I see at a glance is over $400, while I can find Blu-Ray
players with HDMI 1.4 (and oh yeah, an optical drive, video decoder,
etc.) for under $100.

I'm sure some of that price difference is related to manufacturing
volume, but I don't think it is that big of a percentage.

I will say that one nice thing about having different connectors for
different protocols (on consumer devices anyway) is that you don't have
to worry about somebody plugging the Internet into the "Video 1" port
and wondering why they aren't getting a picture.

I do agree but I also think that for HDMI Ethernet your TV (which is the
device with lots of HDMI sockets) will act as an Ethernet switch, so there
shouldn't be any "Ethernet enabled" vs. "Video Enabled" ports.

Now of course that means you probably need Spanning Tree in your domestic


In this day and age exactly how hard is this? Since it's all linux
under the hood, isn't it just a brctl away?


HDMI is also extremely distance limited. At those kinds of distances
you probably would have no problem running 8 gbps over a Cat 6 with
RJ-45s as well. I don't know how many people remember it but 1G used to
be real expensive as well. In a few years you will see the 10 gbps
D-Link switches at Best Buy for $40. Bottom line is that vendor know
that people who need 10G speeds can afford to pay for the privilege.

The important thing about consumer connectors is that plugging a cable
in the wrong place should not blow anything up. You can use an RJ45 for
anything you want as long as plugging that into an Ethernet port or
console port doesn't smoke anything. There is not much magical about an
HDMI cable, it is was just a way for the home entertainment equipment
makers to avoid having your mom hooking up multiple component video,
multichannel audio, and Ethernet and flooding their support phones. For
datacenters there is no such push because there is no telling how many
connections you need to a server and there are geeks like us to figure
out the piles of wires.

Steven Naslund

Distance, data rate required, bandwidth (like RF signals), analog
signals and timing that Ethernet does not provide. I suppose that you
cable box could encode everything as Ethernet/IP to send it to your TV
but it would take lots of processing horsepower to encode/decode. Your
stereo could take the analog output going to your speakers and encode it
as a digital Ethernet/IP signal but then you would need to decode and
amplify it at the speaker. Some signals are better off as analog or RF
end to end. Your FM radio antenna is going to be pretty expensive if
you want to use Ethernet between it and your stereo receiver.

Steven Naslund

I have noticed that too. However it is not the RJ-45 connector's fault.
It is the morons that insist on recessing connectors in places where you
can't get your finger on the tab. I like the patch cords that have the
kind of loop/spring thing for a tab that does not catch on everything
and that way you don't need the boot over the tab. Another pet peeve of
mine is connector boots that harden up over time so it is nearly
impossible to flex the tab to remove the cable. Also, how about the 48
port 6500 blades and trying to remove the cables near the blade
extraction tabs. Grrrr.

Steven Naslund

Some companies such as Apple have completely removed Ethernet ports from
their Pro line laptops.

Other vendors, such as ASUS, have thin laptops with collapsing Ethernet
ports that tuck into the case.

Some companies such as Apple have completely removed Ethernet ports from
their Pro line laptops.

carrying a dongle sucks. but i understand the geometry problem.


this company, but that have a good picture for reference.


The full boots can be so thick that they won't fit into a high-density
switch. If you're in a cold environment they go from difficult to compress
to damn near impossible. More than once I've used a knife to cut a hardened
boot off a cable so it's usable again.


I have seen the sixty or so messages about this and have marveled how
many can major on the minutia and ignore the obvious which Brielle
brings out.

First, Ethernet connectors have changed - Thicknet (RG8) with
transceiver cables, thinnet, and now CAT series cables. Yep, I have
bored in the vampire taps and crimped thinnet.

In another venue I work we still have millions maybe billions of lines
of COBOL code. Why? Because it works. Because the cost of conversion
to something else is prohibitive. It is being done by attrition and I
might say, painfully. One organization I am aware of was to have been
extracted from the tar baby of its COBOL code that was originally
written in 1968 in COBOL D before Y2K had to fix all of that to run
properly over the millennium. And one company I am aware of had to
convert its COBOL F to COBOL II to get there. I haven't followed it
since 2003 but they were still working on getting free from COBOL then
when I was offered a job helping them extricate from the mess. I was
having too much fun with WAN's. BTW, I am retiring 2/28/13 - if anyone
has a COBOL and/or CICS job out there with the right location and
situation I may be interested. I am fantastic as translating COBOL into
a language JAVA coders can understand. I write JAVA, I do not call JAVA
coders programmers. Programming is the next thing to retirement.

And RJ-45 has some of the same characteristics. It works. There are
trillions of them out there in use and on equipment (the corresponding
jacks). There are millions of techs who can put them on. Well, maybe
that is going a little too far. I have seen too many techs who claim to
know how who should be hung with their cabling. They are used for
everything so nearly every wiring discipline knows them. There are
millions of sets of tools to attach them.

I just saw an installation where a ham radio transmitter was set up in a
hospital "in case everything else fails" and they put the transmitter at
the roof, ran a 20 foot pre-made coaxial cable with PL259's to the
antenna and two CAT-5's down to the operator area where they put the
control. The transceiver allows separation of the control head and the
transceiver. The one cat 5 carries the controls - the connectors on the
units are RJ-45. The other CAT-5? They made one pair out of the CAT5,
tied 4 wires together to get enough copper to handle the speaker.
Reason? The hospital wiring staff did not know how to put on a PL259 on
RG-213. (Similar to RG-8). But they could run CAT-5 and put on RJ-45's.

So to change we have to provide training, tools, adapters (another
nightmare), labor to convert and for what? There is no other connector
I am aware of and I haven't heard of any serious contender from anyone
here. That means 30 million dollars development (my estimate) and five
years till we get the beta models. And for what? I can't see any way we
could get more than a 20% higher density, even ignoring noise and
crosstalk issues. And even if we can get 50% more would it be worth it?

Answer, MAYBE in some very specialized and/or badly designed situations
(concentrating too much copper in one place rather than distributing to
"close up switches" with fiber) where a higher density would be of
value, yes. But now we create another set of adapters.

I am a Ham Radio Operator - Extra Class. I work with Emergency
Communications. Having one more connector type is one more big
headache. Yes, if there is a real advantage, fine. Most ham hand held
transceivers went from the venerable and solid BNC to the SMA a few
years ago. They screw a 18 inch antenna on an SMA! Guess what? They
break when you are lucky, otherwise they go intermittent. And just to
make it more interesting one of the Chinese suppliers of very inferior
HT's uses an SMA male on the radio, not an SMA female like everyone
else. So now instead of having three antenna connector types in general
use, N, PL259, BNC, each with their strengths and weaknesses and reasons
to use in certain places, we have 5 with no serous reason for two of
them. Note that HT's have used BNC and SMA, mobiles and bases are
generally N, PL259 with a few BNC. I have standardized on bas/mobile at
PL259 and SMA male for HT to maintain sanity. And to be able to work
with others who have a dukes mixture I carry a small box of adapters.

The IT industry trail is littered with computer languages that were
written to fix some non-existent problem and all that did was create
more confusion. Many claimed to allow anyone to code programs,
something that is true but when you use people who really do not know
how to program you produce tons of shit code that is nasty to make
changes to - and maintenance of programs is usually 90% of life cycle
costs. It is the same in a wire room when you let someone who doesn't
know how to properly place wire do it. PASCAL is one example I can
cite. It had absolutely no advantage over several other languages
existing at the time but academia thought it was cute and pushed it. A
few industries used it and created havoc with it. There were others
like Ideal that many of you never heard of because they were never
widely used but created a conversion opportunity when they were no
longer supported.

Ralph Brandt