Yost Serial Device Wiring Standard (1997) | Hacker News Reader

Yost Serial Device Wiring Standard (1997)

yost.com

26 pointsmmastrac11 comments

I just texted Dave that he made the HN front page, a lifetime achievement to be sure.

My favorite rant of his is web dot.

Interesting take. Considering its from 1997 when serial was still relevant, I'm surprised it didn't really take off.

> I put this scheme together in the summer of 1987. (Thanks to folks at UC Berkeley for the idea. Their wiring is slightly different, I gather for historical reasons, but the basic idea was there. If I can find out who really originated this scheme, I'd like to credit him or her by name.)

I'm pretty sure some of the ISA bus multiport serial cards were using something like this before '87. It was used but not for "consumer" gear.

I am not sure why anyone would use that particular standard.

If your org has lots of serial-using equipment (like routers or port concentrators), it would make sense to use pinout compatible with them. You'd still need to make DCE-side adapters, but at least DTE-side cables could be purchased off-the shelf. And you could use simple straight-through cables, too. The main downside is you might still need null-modem adapters.

If you are designing system from scratch and you really hate null-modem adapters, why require mirror-image wired cables? While they were very common in 4P4C format, the pre-made cables were very rare in 8P8C. Instead, I'd design this using Ethernet crossover pinout -- at least the pre-made cables exist. Oh, and while at it, let's put RX/TX on green and orange pairs -- this way we can use those cheap 2-pair ethernet cables for serial too, if we do not need flow control.

But IMHO, the biggest problem with the standard is this little sentence under pinout:

> The Green and Red wires in the RJ-45 adapter must be soldered or crimped together.

This step is a major pain (before, all you need a single part from amazon + your fingers. For this, you need soldering iron, wire cutters, etc...), and increases assembly time significantly, especially if you have to assemble many converters. Yet it is necessary to fix "hassle 3" and eliminate the need for null-modem adapters.

My preference would be to use straight cables with occasional null-modem adapter. IMHO, the occasional need is not worth it making every single cable harder to assemble.

As someone who wired alot of rs232, I agree that this doesn't make a net reduction in work or confusion.

It even makes one thing worse by using the same rj45 as ethernet, for something other than ethernet. You can (and could in the 80s and 90s) buy idc db9 & db25 which would give you the same crimping convenience while using a connector that is both distinct from ethernet, and actually already pdesent on all the rs232 devices. I did still terminate serial and telcom cables in rj45 patch panels right along side ethernet, but I only considered that a necessity not actually a good thing.

It's maybe kinda sorta neat in the imaginary world where all devices were already wired with right plugs and the right pinout, but not in the world where you have to make all those adapters.

The only situation where it might help would be in some organization that randomly connected modems to modems and then disconnected them and reconnected to printers and terminals, and then swapped around again to some other random connections. What operation ever does that? Maybe a school lab? I never once needed to take a printer that was connected to a computer and connect it instead to a modem or another printer. Everything I ever had to move or change, was only to something equivalent. Maybe moving a printer from a direct connection to inserting a serial ram buffer between the pc & printer, or moving the printer from a direct connection to a print server, or moving a modem from one computer to another computer. The type of cable never needed to change, so a universal cable wouldn't matter.

This ambidextrous thing is kinda nice in theory but really would not have saved me any real time or effort. I'd still have exactly the same effort just in the adapters, only now it's 3 things I have to make and get right and troubleshoot later, instead of one. 2 adapters & a cable instead of a cable.

The part of the system from adpter to adapter is simpler, but the entire system including the adapters is actually worse than the normal way.

At the very least, to give this any justification at all, it has to use a pinout that works with off the shelf ethernet cables, which this pinout doesn't. It does work with flat phone cables, but those suck for long rs232 runs because of no shielding and parallel lines.

I worked in a lab where we invented basically the same scheme in about the same time. I still have a box full of those little grey plastic RJ-45 to DB-9/15/25 M and F adapters.

The cable plant improvement was good - we could punch down CAT-3 (later CAT-5) into keystone jacks and patch panels without regard for whether it was RS-232 or RS-422/485. The big advantage is that you only have one type of wiring and that it's flexible as to whether it was 10Base-T (later 100) or serial.

I used a different pinout that worked well for us:

1. almost all of our systems were 3-wire (no hardware handshake, just RX/TX/GND). In a few cases we needed extra signals and then used the other two pairs.

2. we treated each RS-232 as a pair of GND/signal. Put Rx/GND on 1/2. Tx/GND on 3/6. this provided a little better "shielding" because the ground wire was twisted around the signal wire vs. having TX/RX twisted with each other.

2a. In RS-422/485, used the same pairs but RX+/RX- on 1/2 and TX+/TX- on 3/6 with a chassis ground on one other pin.

3. By using 1/2 and 3/6 it made standard Ethernet cables "straight-through" and made Ethernet cross-over cables into null modems. I wired terminal servers and workstation adapters "backward" (not unlike Ethernet switches vs. hosts or DCE/DTE). This made for very few crossovers required.

This kept our wiring neat and very flexible. You can color code Ethernet vs. serial. The biggest disadvantage was the shitty stamped tin pins in the cheap adapters.

Still used a ton in manufacturing.

I know, I still use it myself too. I just mean it's not relevant anymore in mainstream computing and it was in 1997 when most people still used dialup modems.

Cisco pinout took over, and almost everything went that way.

I certainly use it in my home networking. It works really well in that range of distances

That's true, RS232 has always been very forgiving at least if you use good drivers. Like the 16550.

I've had really bad experience with some MAX232 chips, especially the one that didn't need external capacitors (I forget the model number). Those tend to cause issues and blow up at the smallest wiring mistake.

I still use RS232 today, even for actual computing as I have an actual VT520 terminal <3 And I use it a lot in electronics projects where it is still very widespread.

Reply on news.ycombinator.com