VOGONS


Reply 20 of 46, by maxtherabbit

User metadata
Rank l33t
Rank
l33t
acl wrote on 2025-02-18, 16:50:
Never had first hand experience with BNC networks (i'm a bit too young) But just looking a the "BUS" network topology, the perfo […]
Show full quote

Never had first hand experience with BNC networks (i'm a bit too young)
But just looking a the "BUS" network topology, the performance should decrease rapidly with the number of systems connected.
All the stations are sharing the same ethernet collision domain and CSMA/CD would throttle the packet rate quite a bit.
Am i correct ?

the same problem applies with 10baseT when using hubs, which was standard at the time

the performance hit isn't really that noticeable, at least in my experience with networks consisting of 10-50 stations

Reply 21 of 46, by mkarcher

User metadata
Rank l33t
Rank
l33t
alfiehicks wrote on 2025-02-18, 13:58:

I was already thinking about using a DA-15 extension cable to solve this problem. As long as the wiring is straight-though, I don't see why that wouldn't work.

A 1:1 pinout will work. Be aware that the cable carries the ethernet signal at 10MBit/s, so using twisted pairs for the RX and TX channels is definitely preferred over a simple 15-pin extension cable.

Reply 22 of 46, by the3dfxdude

User metadata
Rank Oldbie
Rank
Oldbie

Things were a bit different in those days. Networks weren't used for streaming and gaming. Internet was in its infancy. Companies could use networks for file sharing, email, printing. And computers were alot slower then. And alot of network computing can be done at modem speeds even. And alot of ethernet alternatives that could do the job. So while ethernet at that time in theory could be alot faster with better topology, and faster computers; really there was room to grow for some time, so we got away with cheaper solutions. Like people say, 3c503 was an early card that worked well enough in PCs at the time to help popularize options in networks, I personally did not get into it that early on, it was just a little too much to run for me at home. But it was around, even if you didn't need to blaze the full 10mbit/s. The deficiencies got apparent later. It's totally understandable why these cards are so handy today, now how prevalent and standard TP is.

Reply 23 of 46, by DaveDDS

User metadata
Rank Oldbie
Rank
Oldbie

The DA-15 network connector was once fairly common .. I think I still have a hub or two
with them...

I was called "thicknet" - and there have been a few discussions on Vogons about it before:

15-pin connector on ethernet card?

Dave ::: https://dunfield.themindfactory.com ::: "Daves Old Computers"->Personal

Dave ::: https://dunfield.themindfactory.com ::: "Daves Old Computers"->Personal

Reply 24 of 46, by dionb

User metadata
Rank l33t++
Rank
l33t++
acl wrote on 2025-02-18, 16:50:
Never had first hand experience with BNC networks (i'm a bit too young) But just looking a the "BUS" network topology, the perfo […]
Show full quote

Never had first hand experience with BNC networks (i'm a bit too young)
But just looking a the "BUS" network topology, the performance should decrease rapidly with the number of systems connected.
All the stations are sharing the same ethernet collision domain and CSMA/CD would throttle the packet rate quite a bit.
Am i correct ?

Correct that Ethernet works like that, but it's not a consequence of the bus topology. Star topology Ethernet has exactly the same issue - everything on the same wire is in the same collision domain. Scaling was awful, which is why bridges were common and people considered things like Token Ring superior.

That only changed once hubs were replaced by switches, which effectively gave every system a dedicated point-to-point connection and collisions were a thing of the past.

Reply 25 of 46, by acl

User metadata
Rank Oldbie
Rank
Oldbie
Jo22 wrote on 2025-02-18, 17:01:
acl wrote on 2025-02-18, 16:50:
Never had first hand experience with BNC networks (i'm a bit too young) But just looking a the "BUS" network topology, the perfo […]
Show full quote

Never had first hand experience with BNC networks (i'm a bit too young)
But just looking a the "BUS" network topology, the performance should decrease rapidly with the number of systems connected.
All the stations are sharing the same ethernet collision domain and CSMA/CD would throttle the packet rate quite a bit.
Am i correct ?

Edit: https://en.wikipedia.org/wiki/PhoneNET
AppleTalk software was also available to IBM PC running DOS/Windows 3.

Never heard of AppleTalk/PhoneNET before thanks for the link

dionb wrote on 2025-02-18, 19:40:

Correct that Ethernet works like that, but it's not a consequence of the bus topology. Star topology Ethernet has exactly the same issue - everything on the same wire is in the same collision domain. Scaling was awful, which is why bridges were common and people considered things like Token Ring superior.

That only changed once hubs were replaced by switches, which effectively gave every system a dedicated point-to-point connection and collisions were a thing of the past.

Things changed when network devices got less expensive.

I'm also a big fan of serial based solution. Instead of a hard to find 8bit Ethernet card, OP could use a serial cable to a raspberrypi + PPP and have the raspberry to route the packets between the PPP connected DOS machine and the home network. I never tried that from DOS, but works well form W95.

"Hello, my friend. Stay awhile and listen..."
My collection (not up to date)

Reply 26 of 46, by chinny22

User metadata
Rank l33t++
Rank
l33t++
maxtherabbit wrote on 2025-02-18, 17:45:

All the stations are sharing the same ethernet collision domain and CSMA/CD would throttle the packet rate quite a bit.
Am i correct ?

the same problem applies with 10baseT when using hubs, which was standard at the time

the performance hit isn't really that noticeable, at least in my experience with networks consisting of 10-50 stations
[/quote]

This is my experience as well. Collisions where common but also expected, but for the gaming side of things it was still faster than modem speeds of the time so didn't really matter.
Multiplayer code was optimised allowing for high latency.
probably the only one not plugged direct to the switch via RJ45 it'll have that segment of the network to itself anyway.

But I agree, there is something cool about having everything networked up, plus file transfers are more continent, if slow

Reply 27 of 46, by alfiehicks

User metadata
Rank Newbie
Rank
Newbie
acl wrote on 2025-02-18, 20:57:

I'm also a big fan of serial based solution. Instead of a hard to find 8bit Ethernet card, OP could use a serial cable to a raspberrypi + PPP and have the raspberry to route the packets between the PPP connected DOS machine and the home network. I never tried that from DOS, but works well form W95.

I do have a Raspberry Pi lying around somewhere, so I could do that. In terms of cabling, would it just be a case of a serial to USB -> Raspberry Pi -> network switch? Although I don't have a serial/parallel combo card, so I'm stuck with DB25 serial ports, so another adaptor would be needed to get DE9, unless you can just buy a DB25 to USB cable? I've never seen one, though.

But I'm even more clueless on the software than I am on the hardware: what's a PPP?

Reply 28 of 46, by acl

User metadata
Rank Oldbie
Rank
Oldbie
alfiehicks wrote on 2025-02-19, 08:18:
acl wrote on 2025-02-18, 20:57:

I'm also a big fan of serial based solution. Instead of a hard to find 8bit Ethernet card, OP could use a serial cable to a raspberrypi + PPP and have the raspberry to route the packets between the PPP connected DOS machine and the home network. I never tried that from DOS, but works well form W95.

I do have a Raspberry Pi lying around somewhere, so I could do that. In terms of cabling, would it just be a case of a serial to USB -> Raspberry Pi -> network switch? Although I don't have a serial/parallel combo card, so I'm stuck with DB25 serial ports, so another adaptor would be needed to get DE9, unless you can just buy a DB25 to USB cable? I've never seen one, though.

But I'm even more clueless on the software than I am on the hardware: what's a PPP?

Yes thats the idea.

Retro PC <-> Null modem serial cable <-> Serial to USB <-> Raspberry <-> home network (RJ45)

On win95 it's set up like a dialup modem connection but with a specific modem driver that allow a direct serial connection instead of having a modem in between (Win98 have that feature included by default iirc). I connect to the raspberry with a dialup config. It is configured with interactive login mode so it opens a shell to authenticate on the Linux system. Once logged i start the PPP command. This takes control of the terminal and sends magic sequence that tells the calling side that the PPP protocol can start. The "modem connection" is initiated and with the proper ip and routing setup on the linux box this allow the Win95 system to access your local lan and even internet

Replace the null modem with a pair of real modems an a telephone line : congratulations, you just built a 90's style ISP 🥳

Of course this can be automated (I keep the manual terminal login step to have more info if something fails)

I might consider writing a guide about that if some people are interested. But I never tried that on dos. Only W95/98

(Edits: bad English and typos)

Last edited by acl on 2025-02-19, 18:48. Edited 1 time in total.

"Hello, my friend. Stay awhile and listen..."
My collection (not up to date)

Reply 29 of 46, by Jo22

User metadata
Rank l33t++
Rank
l33t++

Hi, DreamPi for the Sega Dreamcast works similarly, but uses two modems.
https://www.youtube.com/watch?v=NFUTInM7gq8

"Time, it seems, doesn't flow. For some it's fast, for some it's slow.
In what to one race is no time at all, another race can rise and fall..." - The Minstrel

//My video channel//

Reply 30 of 46, by eisapc

User metadata
Rank Member
Rank
Member

I allways wanted to check if DOS 6.22 Interserve runs on Dosbox or some other virtualized system.
I had good experience with the performance of the parallel connection.
Anybody tried this allready?

Reply 31 of 46, by kaputnik

User metadata
Rank Oldbie
Rank
Oldbie
Jo22 wrote on 2025-02-18, 14:18:
That seems right, I've read about it a few times. Is that 8p8c term beeing used on an international basis or is it US only termi […]
Show full quote
alfiehicks wrote on 2025-02-18, 13:58:
dionb wrote on 2025-02-18, 13:27:

Nit-pick mode: it's a DA-15 connector, not a DB-15. The letter indicates shell size and DB is the size of a 25p serial or parallel connetctor.

Yes, that's true. Further nitpick mode: RJ45 isn't the correct term either, it's 8p8c.

That seems right, I've read about it a few times.
Is that 8p8c term beeing used on an international basis or is it US only terminology?

Because, I remember that the RJ11 connector as used by phone cabling has another, more correct term, as well.
It's not being used were I live, though. RJ11 is exotic enough, already.
The term "western connector" is probably more widespread, at least among older IT people here.

By the way, in my country, FRG, the "RS-232 port" (aka EIA-232 port) often was being described as "V.24 port", as well.
In daily language, in the pre-Windows 95 days, IT people talked about the "V24" when working with serial ports. It even made it into books.

Though strictly speaking, both terms do refer to slightly different things, I think.
One is the protocol side, the other the electrical side, if memory serves.

8P8C (8 positions, 8 contacts) is the correct (international) designation for the modular connector itself. The RJ (Registered Jack) designations refer to the telephony standards using those modular connectors. RJ11 uses 6P4C connectors, note that the outermost positions aren't populated in RJ11 plugs.

The ITU-T V series generally standardizes data communication over telephony networks. I'm not very well read up on it, but it wouldn't surprise me at all if it specifies the whole chain, including connectors.

dionb wrote on 2025-02-18, 15:02:

One pitfall though: these days there's a market for leading Ethernet over 75Ohm (TV) coax passively using baluns. They can have BNC connectors and look very similar to a transceiver but most definitely don't work the same.

Very good point. This is no small niche market either. Being able to use the existing 75ohm coax wiring simplifies and cost reduces upgrades of old analog CCTV networks immensely. Just install new IP cameras and use the old wiring with those media converters instead of tearing down the whole cruise ship/mall/airport/factory/power plant/whatever causing downtime and costs to pull new cables 😀

Reply 32 of 46, by zyga64

User metadata
Rank Oldbie
Rank
Oldbie
dionb wrote on 2025-02-18, 15:02:
Alternatively if you can't find one of those for an acceptable price (they still sell them new for USD 130...) you can look for […]
Show full quote

Alternatively if you can't find one of those for an acceptable price (they still sell them new for USD 130...) you can look for an old 10Mb hub which invariably has a BNC port and so acts as a transceiver:

ER-5390P_L_1000x1000.jpg

One pitfall though: these days there's a market for leading Ethernet over 75Ohm (TV) coax passively using baluns. They can have BNC connectors and look very similar to a transceiver but most definitely don't work the same.

I just bought a 3Com 3C16701 (8x RJ45 ports + 1x BNC 10MBit Ethernet HUB) for about ~$5. However, I haven't tested it yet.

Scamp: 286@20 /4M /CL-GD5422 /CMI8330
Aries: 486DX33 /16M /TGUI9440 /GUS+ALS100+MT32PI
Triton: K6-3+@400 /64M /Rage Pro PCI /ES1370+YMF718
Seattle: P!!!750 /256M /MX440 /Vibra16s+SBLive!
Panther Point: 3470s /8G /GTX750Ti /HDA

Reply 33 of 46, by Jo22

User metadata
Rank l33t++
Rank
l33t++
kaputnik wrote on 2025-02-20, 10:05:
Jo22 wrote on 2025-02-18, 14:18:
That seems right, I've read about it a few times. Is that 8p8c term beeing used on an international basis or is it US only termi […]
Show full quote
alfiehicks wrote on 2025-02-18, 13:58:

Yes, that's true. Further nitpick mode: RJ45 isn't the correct term either, it's 8p8c.

That seems right, I've read about it a few times.
Is that 8p8c term beeing used on an international basis or is it US only terminology?

Because, I remember that the RJ11 connector as used by phone cabling has another, more correct term, as well.
It's not being used were I live, though. RJ11 is exotic enough, already.
The term "western connector" is probably more widespread, at least among older IT people here.

By the way, in my country, FRG, the "RS-232 port" (aka EIA-232 port) often was being described as "V.24 port", as well.
In daily language, in the pre-Windows 95 days, IT people talked about the "V24" when working with serial ports. It even made it into books.

Though strictly speaking, both terms do refer to slightly different things, I think.
One is the protocol side, the other the electrical side, if memory serves.

8P8C (8 positions, 8 contacts) is the correct (international) designation for the modular connector itself. The RJ (Registered Jack) designations refer to the telephony standards using those modular connectors. RJ11 uses 6P4C connectors, note that the outermost positions aren't populated in RJ11 plugs.

Interesting. I'm still not sure if that's a formal thing or actually being used in daily language in countries across the world.

For example, I've just learned that "UTP" means unshielded (yuck) twisted-pair. No, really!
While I'm aware of the nickname "twisted pair" for 10BaseT cabling (and phone cabling), "TP" or "UTP" acronym was never mentioned in the network or computing literature I had read.

I've checked, RJ11 is also known as 6P4C or something?
I've asked my father what "6P4C" is and he shrug, said he never heard of it.
When I asked him about "RJ11", he answered "telephone cables?".
He's been into computing since the 70s (CP/M era) and had installed novell networks in the 80s/early 90s (here in Europe).

"Time, it seems, doesn't flow. For some it's fast, for some it's slow.
In what to one race is no time at all, another race can rise and fall..." - The Minstrel

//My video channel//

Reply 34 of 46, by davidrg

User metadata
Rank Member
Rank
Member
eisapc wrote on 2025-02-20, 07:53:

I allways wanted to check if DOS 6.22 Interserve runs on Dosbox or some other virtualized system.
I had good experience with the performance of the parallel connection.
Anybody tried this allready?

Whether it does or not, it feels like it wouldn't really be worth the effort. Null-modem cables are easier to get and don't require getting Dosbox involved if you just use Kermit which is available natively on DOS and Windows (disclaimer: I'm the Kermit 95 maintainer) and everything else under the sun for free. If serial is too slow, just use it to transfer over a network driver!

alfiehicks wrote on 2025-02-18, 12:36:

Okay, so I know next to nothing about networking, but I have so far managed to bumble my way into getting all of my old machines connected to my home network. There's no real practical reason for this, as they're all sneakernet-able in various ways, but it's cool.

But there is a practical reason for this! Once its all setup its easier and faster! While I can sneakernet all my PCs, if they can take a NIC they get a NIC and they go on the LAN. With network drives it gives them practically infinite disk space and I don't have to waste time copying files around. I used NetWare for this, but mTCP NetDrive and EtherDFS are likely easier options if you're currently just FTPing or sneakernetting things around.

Reply 35 of 46, by kaputnik

User metadata
Rank Oldbie
Rank
Oldbie
Jo22 wrote on 2025-02-20, 18:30:
Interesting. I'm still not sure if that's a formal thing or actually being used in daily language in countries across the world. […]
Show full quote
kaputnik wrote on 2025-02-20, 10:05:
Jo22 wrote on 2025-02-18, 14:18:
That seems right, I've read about it a few times. Is that 8p8c term beeing used on an international basis or is it US only termi […]
Show full quote

That seems right, I've read about it a few times.
Is that 8p8c term beeing used on an international basis or is it US only terminology?

Because, I remember that the RJ11 connector as used by phone cabling has another, more correct term, as well.
It's not being used were I live, though. RJ11 is exotic enough, already.
The term "western connector" is probably more widespread, at least among older IT people here.

By the way, in my country, FRG, the "RS-232 port" (aka EIA-232 port) often was being described as "V.24 port", as well.
In daily language, in the pre-Windows 95 days, IT people talked about the "V24" when working with serial ports. It even made it into books.

Though strictly speaking, both terms do refer to slightly different things, I think.
One is the protocol side, the other the electrical side, if memory serves.

8P8C (8 positions, 8 contacts) is the correct (international) designation for the modular connector itself. The RJ (Registered Jack) designations refer to the telephony standards using those modular connectors. RJ11 uses 6P4C connectors, note that the outermost positions aren't populated in RJ11 plugs.

Interesting. I'm still not sure if that's a formal thing or actually being used in daily language in countries across the world.

For example, I've just learned that "UTP" means unshielded (yuck) twisted-pair. No, really!
While I'm aware of the nickname "twisted pair" for 10BaseT cabling (and phone cabling), "TP" or "UTP" acronym was never mentioned in the network or computing literature I had read.

I've checked, RJ11 is also known as 6P4C or something?
I've asked my father what "6P4C" is and he shrug, said he never heard of it.
When I asked him about "RJ11", he answered "telephone cables?".
He's been into computing since the 70s (CP/M era) and had installed novell networks in the 80s/early 90s (here in Europe).

If we're going to engage in some nitpicking, the standard designation is U/UTP, where the first U means that there's no overall screen, and the second one means the wire pairs aren't individually screened either. It's often abbreviated to UTP though. Guess you've figured out that S and F means (braid) screened and foiled respectively too 😀

When it comes to TP cables in general, if you ask for one without further specification, you'll get a XBaseT one. In practice, there are TP cables of any imaginable type and configuration though. They're useful for almost any kind of high frequency signalling, since they enable a very simple way to reduce EMI/crosstalk. Send the signal in one wire of the pair, and the inverse of the same signal in the other. Their induced magnetic fields will of course also be each other's inverses and cancel each other out.

Yeah, it's kinda funny, those RJ designations are so universally used that almost no one knows the correct ones here either. Sometimes you see electronics retailers use them though, often within parentheses, "RJ45 modular jack (8P8C)" 😀

Reply 36 of 46, by akimmet

User metadata
Rank Member
Rank
Member

A "real" RJ45 connector is keyed, it will not fit an Ethernet jack. It was a common ordering mistake long ago. In fact there is more than one standard for contact spacing, making things even more confusing if you are crimping your own cables.
https://www.arrow.com/en/research-and-events/ … rj45-connectors
https://en.wikipedia.org/wiki/Modular_connector

There is a reason UTP won out over coaxial cabling.

Coaxial cable has significantly more capacitance, making long runs difficult to drive. Coax is also more susceptible to signal reflections and ground loops. While 10base2 is supposed to be galvanicly isolated, in practice it isn't. The t-connectors and terminator resistors are uninsulated and can touch multiple grounding points.

Twisted pair balanced signaling has far less signal loss over distance. Another bonus is since twisted pair uses differential signaling, it has some noise immunity. Noise will be picked up in phase on both wires, which will end up being canceled out at the receiving end.
https://en.wikipedia.org/wiki/Balanced_line

Reply 37 of 46, by Jo22

User metadata
Rank l33t++
Rank
l33t++
akimmet wrote on 2025-02-21, 05:05:

A "real" RJ45 connector is keyed, it will not fit an Ethernet jack. It was a common ordering mistake long ago. In fact there is more than one standard for contact spacing, making things even more confusing if you are crimping your own cables.

Ha! That reminds me of our TAE phone connectors, too! 🙂
They had TAE-N (answering machine, fax/modem) and TAE-F (telephone) connectors, which had a different "nose".
Cutting it was a common "hack" to use the wrong connector, if the correct one wasn't available.
https://en.wikipedia.org/wiki/TAE_connector#TAE_socket

Edit:

There is a reason UTP won out over coaxial cabling.

Money. far as the US is concerned, I can hardly imagine another reason. It's always money.
Anyhow - from a neutral point of view coaxial cable is more precious and of higher quality than bell wire.
That's what cable TV cabling uses, too. It outperforms DSL (twisted-pair cabling) in terms of bandwidth.

That being said, optical connections are the future - since the 1970s.
Our postal minister here once tried to establish fibre, but the money went into cable TV.
Because stupid television was considered more important than telecommunications, or so the officials thought. Sigh.

Also sad is that ISDN (introduced late 80s/early 90) already had fibre cabling, but it got replaced by DSL and twisted-pair (ew) cabling by ca. 2000.
Pro DSL people claimed that the ISDN fibre was being too limited anyway,
but it wasn't taken into account that the optical transceivers were modular and replaceable.
The deployed fibres as such were quite capable. Anyhow, these are just very old stories.. 😅

I hope that optical connections will finally make it this time, maybe even into LANs. 🙂
The lack of radio interference and ground loops would be quite nice for a change.
I'm optimistic. The insane bandwidth requirements these days will be an, err, "inspiration". 🙂

Twisted pair balanced signaling has far less signal loss over distance. Another bonus is since twisted pair uses differential signaling, it has some noise immunity.
Noise will be picked up in phase on both wires, which will end up being canceled out at the receiving end.

Nice theory. However, in practice, DSL and shortwave radio do cause interference with each others. 🙁
Go have a look at amateur radio forums. DSL and RJ45 cables are often cause of noise below 30 MHz, despite their twisted-pair cabling.

Edit: Some learning/info material:
http://www.arrl.org/dsl-interference
https://www.iol.unh.edu/sites/default/files/k … dsl/RFI_DSL.pdf
https://documentation.extremenetworks.com/npb … 39047DE76.shtml

(If almighty twisted-pair would live up to its reputation, then this issue wouldn't exist.)

Adding a couple of big snap ferrites to cables (incl. DC cables) sometimes helps as a workaround to reduce noise.

Edit: An example I've found online: http://alloutput.com/amateur-radio/ethernet-r … oise-reduction/

Edited. 4x.

"Time, it seems, doesn't flow. For some it's fast, for some it's slow.
In what to one race is no time at all, another race can rise and fall..." - The Minstrel

//My video channel//

Reply 38 of 46, by akimmet

User metadata
Rank Member
Rank
Member

There are tradeoffs to both methods of signaling. In the case of ethernet, UTP is clearly the better choice.

DSL is an excellent example of how not to do things. Implementing high frequency signaling over cabling that was barely adequate for telephone use.

DSL was never as insane as Broadband over power line. Thankfully that never caught on. Despite the startup companies claiming it was the greatest idea ever

Reply 39 of 46, by Jo22

User metadata
Rank l33t++
Rank
l33t++

@akkimet Thanks! Yeah, I remember Devolo's earlier PLC technology, it scared the ham community. 😨
By comparison, the issue with twisted-pair was really calm.

acl wrote on 2025-02-18, 16:50:

Never had first hand experience with BNC networks (i'm a bit too young)
But just looking a the "BUS" network topology, the performance should decrease rapidly with the number of systems connected. [..]

Hi again, I had a look at this issue (collisions etc) and it seems that other network technologies such as ARCnet had used a "Token" that's been passed from on station to the next.

https://en.wikipedia.org/wiki/Token_passing

In early 90s, the ARCnet cards for PC looked just like BNC ethernet cards,
with biggest visual difference being the presence of DIP switches sometimes.
They could even use the RG58 cables, albeit the wiring was a bit different.
Star-like, more like what twisted-pair uses.

https://en.wikipedia.org/wiki/ARCNET#Description

Anyway, just saying. ARCnet came to mind ,because I've seen quite some ARCnet cards on eBay recently when I was looking for NE2000 compatible network cards for ISA slot.

Edit: Interesting. There were (are) even PCIe versions being made.
Example: https://radictech.net/products/arcnet/arcnet-cards-2/

Edited.

"Time, it seems, doesn't flow. For some it's fast, for some it's slow.
In what to one race is no time at all, another race can rise and fall..." - The Minstrel

//My video channel//