Re: Raspberry Pi 4's IP Address 169...

Am Sonntag, 09. Januar 2022, um 21:01:33 Uhr schrieb NY:

What about the encryption? Maybe that is the reason for that, but I don't know much about it. With Open-System you should be able to see all the traffic.

Am Sonntag, 09. Januar 2022, um 20:59:42 Uhr schrieb Richard Falken:

That works only for traffic that needs to be routed.

True, because IPv6 uses the neighbor discovery protocol with neighbor solicitations and neighbor advertisements. That is the counterpart for IPv4's ARP.

It has the same "problems" - it can also be spoofed.

Also you can pass a *lot* more data around the network with switches - the effective bandwidth of switch fabric is enormous.

Am Dienstag, 11. Januar 2022, um 13:14:10 Uhr schrieb NY:

It is possible, there are media converters for AUI to 10Base2 and for AUI to 10BaseT. There are also hubs that support all 3 types.

Yes, it is. All computers receive all data and only half duplex is allowed.

Am Dienstag, 11. Januar 2022, um 14:20:13 Uhr schrieb The Natural Philosopher:

IIRC the collision detection checks the RX pair when in half duplex mode, so the change of a collision doesn't decreases.

Am Sonntag, 09. Januar 2022, um 15:00:57 Uhr schrieb The Natural Philosopher:

If you use /etc/network/interfaces you need to create a separate entry for the new interface. You need to know its name (ip a tells you) and the set it up like the other interface, just replace the name (e.g. eth1 instead of eth0).

There are inexpensive managed switches (around 40 EUR), e.g. Zyxel GS1200 series. They have a feature called 'monitor port' which allows assigning one of the switch ports to listen for traffic on selected other ports.

A good hint for the managed features is that the switch is specified 'VLAN compatible'.

That takes me back: vampire taps (which could only be inserted at marked places along the coax, presumably where nulls/maxima were); large transceiver attached to vampire tap; very thick drop cable with sliding locks on D connectors at each end, between transceiver and computer's network card. Moving a LAN cable or a drop cable was a bit was like wrestling a snake ;-) And that was as recent as 1990.

Even the transition from cylindrical Cat 5+ to flat Cat 5+ cable was quite revolutionary when it came out a few years ago. Now it was possible to fit a cable underneath a carpet (ideally close to the edge, between the gripper-rod and the edge of the carpet) rather than having to tuck it down the gap between carpet and skirting board. Also possible to lay it underneath the metal strip between one carpet and another in a doorway. For the first time it was possible to run several Cat 5+ cables side by side, when the gap between carpet and skirting board would only fit one cylindrical cable.

Thank goodness that modern Ethernet ports are auto-sensing and (AFAIK) all cables are straight through. No more being caught out by devices which did or didn't want a crossover cable.

....and the two cat 5 wiring standards have become essentially one...

What is impressive, is that the Ethernet standard has kept pace with all this to gigabit and possibly beyond

Well beyond, there are standards for 10, 40, 100 and 400 gigabit over fibre also 10 and 40 gigabit over twisted pair(s), cat 8 for 40 gigabit so no reusing the old cat 5 wiring but still impressive. There's work in progress on 800 gigabit and 1.6 terabit standards.

It seems Moore has moved his law to network bandwidth after getting bored with processor speeds.

:0-)

Yup. I was pondering the first modem I saw around 1972. 50 baud and an acoustic coupler. By 1982 ir was 300 baud By 1992 we were happy at 9600... at 2002 I had migrated via 64k ISDN to ADSL 256bps fixed rate. by 2012 I was ion ADSL 2 at 6Mbps By 2022 I am on 40Mbps on fibre.

I am still on 100Mbps ethernet, internally because for my little home setup its 'fast enough' and the 20 year old cable run its reliably

On Wed, 12 Jan 2022 14:11:31 +0000, Ahem A Rivet's Shot snipped-for-privacy@eircom.net declaimed the following:

Processor /speeds/ seem to have maxed out -- instead they are adding more and more parallel cores to the processor.

Hear, hear—and thus it has been for decades!

At a presentation in the 1990s, after diagramming the progress of Moore’s “Law”, an attendee remarked that “it was a shame that there was no equivalent law for fibre bandwidth.”

My answer was that there was an exponential increase in bandwidth, even over existing fibre, since the transmitting and receiving technology was driving it.

It is remarkable when any technology standard has a useful life over a decade, and the Ethernet family has proven very adaptable and resilient.

I'm pretty sure the first one I saw in 1975 was 300 baud on an acoustic coupler - it could keep up with an ASR-33 reader/punch.

About that time 14,400 came in, then 28.8, 33.6 quite rapidly and then the asymmetric standards with a digital signal on one end and a modem on the other (V90, V92).

I went from ISDN to 6Mb/s fixed wireless DOCSIS in this period.

Then 70Mb/s LTE around 2016 before landing on my current gigabit FTTP.

I wired this place eight years ago, so it got CAT-6 and a cheap 24 port managed gigabit switch, there's a fighting chance I might get 10 gigabit to run if I feel like spending money madly - there's no point.

the term you are looking for is port mirroring, & can be configured on a managed switch. unfortunately not possible on consumer grade switches as they are invariably unmanned

There's no such thing as fast enough when it comes to networking!

If it's CAT5e you should be able to run gigabit over it without problems.

CAT5 which is fine for 100MB may light up the gigabit LED on a switch, but the throughput maybe terrible if the cables aren't up to it, or have been kinked during installation.

---druck

I suppose with network bandwidth we are close to reaching the limit of the rate at which *one computer* can send or receive over a network - because of current limitations in other parts of a computer: CPU, bus, HDD/SSD. But we now have communications which can carry *many* such conversations from many different computers all talking at the same time.

I can see the difference that network communications makes when I changed from a Pi 3B+ to a Pi4. The Pi 3 has a gigabit Ethernet port which is throttled by the fact that it is connected to the CPU by USB. With an external USB-connected HDD shared using Samba, a Windows computer can copy large 1 GB (*) files to/from the Pi's share at about 130 Mbps maximum. The Pi 4, with the same setup, can manage about 800 Mbps, so not far short of the 1 Gbps limit. Of course some of that increase may be due to the Pi 4 having USB3 rather than just USB2, though even USB2 is 400 Mbps.

Then we have wifi. My old router was wireless N on 2.4 GHz. I was lucky to get anywhere *near* that speed; usually a large file transferred at under

It is getting to the stage where a large video file can be accessed (eg when scrolling quickly through it or moving from one part of the file to another (**)) across an Ethernet or wifi network almost as quickly as if the file was on a local disk connected by SATA. We're not there yet, but it won't be long before it's difficult to tell.

(*) For example a recorded TV programme as a TS file - typically about 1.3 GB/hour when recorded from a BBC channel, a bit less for ITV.

(**) I use VideoReDo to edit the continuity announcements and commercials from programmes I've recorded. I need to shuttle quickly through a file to find each of the commercial breaks. Leaving aside the fact that SD recordings (MPEG2) are a *lot* quicker to process than HD (H264) when shuttling through, I can see a bit of difference between accessing a local file and one on the Pi over Ethernet, and a significant difference if the Windows computer uses wifi instead of Ethernet. But technology will improve, I'm sure.

Yes I remember the old 300 baud acoustic coupler: a polished wooden box, lined with green baize with mic/speaker "holes" which exactly matched the handset of a standard GPO/BT telephone. That was the comms between the teletype that we had at school and the mainframe computer that we accessed - in the late 70s. The teletype was painfully slow, but sometimes we could see it stall for a second if the comms couldn't keep up.

Yes I remember a friend had a standard 28.8 kbps modem for accessing the internet by dialup. I bought my modem a bit later than him, and 33.6 became possible. I think my modem was US Robotics. I've got a feeling that a PROM upgrade became available that allowed it to work a little bit faster than that, as long as the other end could use the correct protocol.

Then "broadband" started to be the word on everyone's lips. There was talk of some large towns having their exchanges upgraded to support ADSL, but those in villages needed to get more than a certain "trigger level" of firm orders (via ISPs) before BT would even consider upgrading them. In the meantime, a group of us on my housing estate investigated a long-range microwave link to a mast on a nearby hill: companies were setting up in business to supply fast internet this way for places that had not yet reached their BT trigger.

Suddenly came the announcement we never thought we'd hear: BT were abandoning the concept of trigger levels and were committing to add ADSL to every exchange. I lived about 200 metres from my exchange so I knew that I'd be able to get up to 8 Mbps. Initially ISPs (because of BT) were charging various amounts depending on what speed you wanted - or could achieve due to line length. I initially opted for 2 Mbps down and 448 k bps up, because 8 Mbps was a lot more expensive. But the price differential got smaller and smaller, so I upgraded to 8 Mbps. Still only 448 kbps upload, though :-(

That situation lasted until "fibre". We got FTTC and managed about 15 Mbps down and about 5 bps up. Not a huge increase in download speed, but a tremendous increase in upload speed when FTPing files to a server.

At our present house we get about 25-35 down and about 8-10 up. The sync speed fluctuates, taking roughly two weeks to go between one extreme and the other. I'm sure if we had brand new wiring from where the drop cable terminates at the old GPO lozenge box on the gable end, to the socket where the router is plugged in, we'd get a slightly better speed. But it would be a major job involving crawl boards on the roof to get at that socket and to run new cable (with no junctions along the way) to the living room - and it's difficult to know where the demarcation is between BT and owner cabling. Maybe it's at the first BT socket rather than the GPO box.

It looks as if our village is being cabled up for FTTP, as part of BT's phasing out of copper connections. That will cause some interesting problems because the point where the BT cable comes into the house (and therefore where they'd *probably* bring the fibre) has no mains socket near and no way of running a mains cable without taking up a carpet (which means moving a lot of furniture) to dig a channel in the concrete floor to run a spur from the nearest socket on the ring main. And you *need* mains for the fibre-to-Ethernet converter (I've forgotten what BT call it).

They cant shrink the gate size any more. That tends to be what limits speed.

And parallel cores are not the be all, for single thread performance.