_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

Predictably, some naysayer posted a comment about not understanding why anyone would want to fool around with 30+ year old computer technology.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

I find myself wishing that there was some sort of "Like this post" button.
I agree with those last two paragraphs in particular. Firefox does seem to have a problem with efficiency, especially on Windows(which is rather poor itself). And OSs being so inefficient is getting less and less acceptable for me. Windows takes stupidly long to start, and is generally sluggish, which is why I avoid using it. Linux is a lot better, but it could probably be faster as well.

I hope that what I write for my 6502 build will be quick, if not blazing fast. But then, that's a different league altogether; a far easier target.

You're lucky you never used an Amiga for any long time. Windows, Linux, and every other modern GUI operating system is just intolerable to me.

I wonder what one does with a Wi-Fi modem on a C64. Presumably it works on the VIC-20 as well. It could probably be made to work with the PET, or anything else with a VIA or two.

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

I used a "modern computer" for the first time last night since probably about a year ago, the last time I filed my taxes. It's a RPi, couldn't tell you much about it other than that it's plugged into a seldom-used TV, and has a bluetooth keyboard. It runs Raspbian only because the last time I installed NetBSD on one, the boot menu got hosed and I never did get it working quite right again. So for my limited uses, I'll just use Raspbian, as much as I detest Linux.

My daily desktop machine is an Amiga 2000. That's got a wired ethernet connection. But my internet connection is by 4GLTE. I have a 'hotspot' gateway connected to my extensive LANs.

I spend most of my online time with this 'phablet' and a small 'phone'. This thing is meant to be a giant phone, but I have it in a case with a keyboard, so it works like a miniature laptop. It runs some klunky version of Android. The 'phone' is several generations old and runs a different klunky Android.

I use my C64 for a lot of things, too, but don't imagine I have any need to use anything on the internet from it. The Wi-Fi modem might be an interesting solution for data transfer, for which I currently use RS232, but mostly do the 'sneakernet' thing carrying floppy disks round the house.

The modern web browser is a frighteningly complex piece of software. Be amazed that they work at all.

But I certainly share the frustration folks have with them, when you have all of this computing horsepower and it stalls out doing something as mundane as typing in text.

I've been fighting stalls at work. I have a simple HTTP endpoint that normally completes in 20-50ms, but you look at the logs of request times and you'll see a happy string of 50ms then a hard stall of 20s. Finally pinned it down to the I/O handling in the kernel -- it was never my code. Under heavy load, the system simply freezes. Good times.

There's certainly a novelty of running a web server on a C64. But I recall my friend getting a 5MB hard drive for his Apple ][, and the was impetus enough for him to host a BBS that we used for our Mac User group for a short while. Not quite state of the art, but capable. 1200 baud modem ftw.

But then, I look at something like the Raspberry Pi, how cheap it is, how small it is, and how hard the Linux they run on it stresses the machine. And then I look back at when we used to run NeXTStep, on a 25MHz '040, w/20MB of RAM. How that would sing on something like the Pi today. All that space, all those GHz. And then I think how a modern browser would just kill it.

In theoretical computer science, any user interface is considered an application, not an operating system component. Both the command line shell (BASH) and GUI running on Linux are user-space applications, using documented entry points in the kernel API to do their magic.

Microsoft Windows up through Windows for Workgroups 3.11 followed the same model to some extent, mostly making MS-DOS API calls for handling files, etc. However, Windows was also directly touching hardware and for a variety of reasons (buggy drivers being on of them), such direct access could result in total machine fatality, causing all work in process to be lost and in some cases, producing significant filesystem damage. When Microsoft released Windows 95 they welded the GUI to the kernel, working on the theory that if the two were part of the same code base bugs would be less likely to creap in. It didn't quite work that way. Windows 95 and to a lesser extent, Windows 95 SR2, was a crash-prone environment. The present Windows model has retreated from what was present in Windows 95, which means that, for example, the obnoxious user interface characteristics of Vista and later versions can be eliminated with a different shell.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

Truer words were seldom spoken. I "love" the Internet, but port 80 is a technical jungle full of huge snakes, the deepest pits ever dug by man, and landmines. The whole lot needs ripping up and starting again, but of course that'll never happen. I certainly would never have got into computer engineering if my first exposure to it was modern JavaScript frameworks and associated horrors.



I never had the pleasure But at Uni (circa '95) they had HP-UX machines with PA-RISC CPUs and a massive 128MB of RAM. As well as being used for ECAD functions, they each had 2 or 3 X servers hanging off them, ran various services like SMTP and the department web server. And they had uptimes in the range of years...

_________________
8 bit fun and games: https://www.aslak.net/

In 1995, our Pr1me and Pr1me clone minis were finally taken offline (could be plus or minus a year or two) as their entire workload had shifted to a half-dozen or so Sparcstation-10 and -20s. They were running CADDS, some users running locally with the framebuffer, and some on graphic terminals. Several of them were intentionally running internet servers, web and email. Each one was directly accessible on the internet, we had 64516 internet IP addresses!

In 2002, we started to migrate to Windows PCs. The clock speed was 5-20 times those Sparcs (we had an array of different CPU configurations) but the computer was limited to one user and really only one task. Doing anything in the background brought CAD/CAM work to a halt. Doing that one task, they were marginally faster than doing that same task on a Sparc, which had who knows what going on in the background.

I remember installing 64Mb of RAM in a Sparcstation-10 around 1990, and being in awe of how much memory that was.

The very first UNIX server I built, which was in 1988 with an Intel 80386 running at 25 MHz, had a whopping 4MB of RAM, which was major memory for PC hardware at the time. The box also had a SCSI subsystem, consisting of a 136MB (not GB) disk and a SCSI QIC-2 tape drive. I loaded SCO UNIX-386 on this machine and it ran great, all the while able to support dozens of concurrent tasks. I ran that box until 1996, when I replaced it with a more powerful unit running an AMD 486-DX4 (120 MHz) MPU, with 64MB of RAM.

Our current office file and print server runs on an AMD Opteron Barcelona MPU (eight cores) with 32GB of RAM. It has SCSI ultra-320 hardware (total of 600GB disk storage) and is orders of magnitude more powerful than that 80386 machine I cobbled together nearly 30 years ago. So there is some improvement...

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

I think the lesson more so today is not so much how much more powerful the machines are today, but, rather, much more slower the software is.

Granted, modern machines, and modern software can do some amazing things. And I certainly enjoy coding in the higher level languages and systems, as well as enjoy brute forcing through millions of rows of data when I need too.

But that doesn't mean it doesn't give me pause to look back at some of the older systems we wrote, the machines we used, the large user bases they supported. That why I gasp at the sheer power of something like the Raspberry Pi, yet, how, apparently, little you can do with it today.

My largest old system was a 50Mhz 68030 machine, the Alpha Micro AM-3000. We ran 150 users on that machine. All running programs written in BASIC, against an ISAM database. The true power of that machine was the I/O hardware to handle all of the users terminal connections. AMOS, the OS, was crude. But for driving business apps like we wrote, it met the need.

I, too, replaced lots of old machines with SPARC pizza boxes, or other RISC towers. See a SPARCStation plugged in to a terminal concentrator far larger than it was. Green screen, 19.2K terminals. Lot of stuff got done on those old, creaky, things. A lot of stuff still gets done on those old, creaky things.

There is no longer an incentive to writing efficient code. Not helping is that the languages being used these days are veritable resource hogs compared to what we used 25-30 years ago.


A top fuel dragster has enormous power as well, yet is essentially useless as a means of transportation—especially if you want to go more than a quarter mile. I see gadgets such as the Raspberry Pi as just that: gadgets. One of them may be able to execute millions more instructions per second than my POC V2 unit. However, I have my doubts that the Pi is more useful.


I recall that machine. It ran a form of SMC BASIC, if memory serves me, which was one of the two major timesharing BASIC families that descended from the minis of the 1970s. The ISAM database engine in the BASIC interpreter was very efficient, usually able to complete a record lookup in a couple of milliseconds. That was quite impressive for a non-RISC machine at the time.

One of our two Linux boxes has Thoroughbred's Dictionary-IV installed, as we support several clients who run their businesses on that environment. This Linux machine was built out of recycled parts and is powered by two AMD Opteron 246 MPUs, which was new hardware back in 2006. Thoroughbred can do an end-to-end search of an inventory database of 7000+ records in about three seconds, an average of 2300+ records per second, each record being about 3KB data. The BogoMIPS for the system is a hair under 8000. For comparison, our file and print server, which runs on AMD Opteron multi-core technology, has a BogoMIPS of 22,500 with a single MPU. I have a second MPU sitting on the shelf and if I remember some day to find a socket-F MPU cooler, will install it just for giggles.


Up until 2006, every UNIX system we installed connected to a gaggle of terminals (mostly WYSE 60s). In 1997, we started using Equinox's SuperSerial (SST) hardware in our UNIX (and later Linux) servers. Unlike the other concentrators we had used in the past, the Equinox SST hardware had a lot of intelligence and did almost all of the the grunt work in processing serial I/O. The volume of interrupts being generated greatly decreased and as a result, we saw significant performance improvements when a lot of users were busy.

The largest such system we installed was in 1997 at Graham Paint & Varnish Inc. in Chicago. The server was powered by an AMD K6-2 MPU running at a screaming 300 MHz, with 64MB RAM and an all-SCSI storage subsystem (even the CD-ROM was SCSI). We used Equinox's SST hardware, with four 32-port panels attached to the controller card in the server. A total of 102 terminals, 16 (serial) printers and two modems were attached. As this was a factory with a lot of floor space, we used up a huge amount of CAT5 UTP to wire up everything—I seem to recall that we went through nine or ten 1000 foot boxes by the time we were done.

The system worked great and when my then-partner and I finished the installation and got everyone up and running on the new hardware, the company threw a "new computer system" party. Of course, if you had seen the ancient Basic 4 mini that it replaced, you too would have partied. The gal who handled accounts receivable was astonished that a trial balance could now be computed in three seconds, instead of three hours.

The largest system we currently have in service running the Thoroughbred Dictionary-IV environment has 35-40 concurrent users, all of whom who use Dynacomm terminal emulation software to get access. That machine also runs Samba and typically supports 100+ Windows client connections. The server doesn't even work up a sweat with that load (it's powered by an eight core Opteron MPU, total BogoMIPS being about 32,000). Despite several attempts over the years to modernize their software and get away from Thoroughbred they have stuck with it because it works, is very fast and very stable.

BTW, the most powerful server we've shipped to date, as measured in BogoMIPS, has a 16-core Opteron, for a total of 67,000 BogoMIPS, along with serial attached SCSI (SAS) for storage. It would easily support 400+ users.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

I've found the Pi3 to be very useful, for specific purposes. In my dual locations I have some more-or-less weird printers located in other rooms. A Pi3 solved the printing problem in a plug-and-play way. I just connected the Pi3 to a printer via USB, connected power, had the micro-SD inserted w/Raspbian, and after a quick setup of Cups I have the printers available from my PC elsewhere in the building. The Pi3 has wi-fi support, and that's what makes it so easy. In addition to the ability of Cups to be tweaked to handle "non-standard" (no PS, no PDF) printers. It was completely painless (unlike with most if not all of the other Pi-sized/priced boards out there). Write a Raspbian image to SD, and boot. A cheap, reliable solution to a practical problem.
I'll be connecting another Pi3 to my old minicomputer as well, to provide remote connection over wifi.