All of these people who keep saying that webapps can replace desktop applications were simply never desktop power users. They don’t know what they don’t know.
Yeah it would be nice if the web accessibility guidelines also focused on actually using the thing normally. For example: offsetting the scrollbar from the right edge of the screen by 1px should be punishable by death.
I’m currently travelling in China and the total absence of graffiti and the wide availability of public toilets as well as the general cleanliness of the place is a stark contrast to London. I am somewhat dreading that part when I return.
If by tourist track you mean several cities each with a population count greater than London then sure. :) But it is true I was there as a tourist to see a few places and China is an absolutely huge place.
I wish the touristy places in the UK were as clean though.
Also, writing this comment on a crowded, dirty and smelly train from Gatwick Airport, which has 2 carriages inaccessible because of broken doors and crawling due to signal failure on the line, I already miss the clean and comfortable Chinese trains I was taking from place to place.
That said I do get to work 945 here for more money so I am grateful for this for sure. :)
> It will not be caught in development testing — who runs a test for 50 days?
You don't have to run the system for 50 days. You can simulate the environment and tick the clock faster. Many high reliability systems are tested this way.
IIRC the initial value for the jiffies time counter in Linux kernel is initialized at boot time to something like five minutes before the wraparound point, precisely to catch this kind of issues.
It uses a hardware clock, one that pauses during sleep. There is no tick.
If you wanted to see how time impacts the program, you'd prob change fns like calculate_tcp_clock to take uptime as an argument so that you could sanity check it.
Yeah, but if you're going to go through all that work because you have to foresight to test how clocks impact code, it would be simpler to stub and test the few functions that call the code (where the bug was). The bug wasn't in the hardware.
If the U.S. implements a draft, will women be required to register? Right now, only men are required to register with the Selected Service when they turn 18. Would Congress amend the Military Selective Service Act?
An idea I have been thinking about: Increasingly powerful chatbots provide more teaching capacity. I think this will lead to counterintuitive outcomes like teaching environments where the human student is not allowed to use any device but is being taught and tested by a synthetic intelligence.
That would have sadly be an improvement for the majority of teachers I've had in my life, even university classes because they were taught by a TA phd student I couldn't understand.
I never said that AI should replace all teachers. AI can help provide more teaching labor (more teaching capacity). Every child can have a personalized tutor in addition to human teachers.
What paid subscriptions respect the readers? I would love to pay for news from organizations that only get money from readers. For example, I have been paying for The Economist for decades and still see advertisements.
I am surprised by the assumption that each box could only handle one modem. I seem to remember that some DOS BBS packages could handle multiple modems/users concurrently and only needed multitasking operating systems for “door” programs. Am I misremembering?
A guy who was local to me, when I was a kid, wrote multi-user BBS system (called "MUBBS" originally-- I don't remember what the name was changed to later) in Turbo Pascal that had a preemptive multitasking loop running in x86 real mode to handle multiple lines simultaneously. The coolest part was the console was just a "line" so you could logon to the board and interact while somebody was online with the BBS, too. Most other DOS BBS packages were only available for the SYSOP or the caller individually.
Edit: Ugh... I'm gonna have to go back to floppy images to find it. There's a "MUBBS" for Mac from 1992 showing up in search engine results but that's not the one I'm thinking of. It was more like 1989 or 1990.
“as modems got faster, supporting 16 modems on a single machine became impossible, and it was often cheaper to buy a new commodity desktop PC rather than a much more expensive machine with a 16-port serial card capable of handling the IO.”
You are correct, but the problem was the PC only had 16 IRQs. That required using intelligent multi-port cards from Digi or Rocketport. They worked by aggregating all the ports to a single card IRQ, and managing all the hardware signals, echo.
I wrote the software for a breakout box that could handle 128 serial ports. It was an ISA backplane with an industrial 286 computer and multi-port serial cards. This was our solution for a MajorBBS system.
The BBS software would have to timeslice between all the cards handling each IRQ, then poll the card details to see which ports needed service.
GalactiComm eventually came out with their own around 1993 that could go out to 255 serial ports and did not require the 286 processor.
By the mid-90’s, Livingston PortMasters were the preferred way to aggregate serial connections, which quickly gave way to USR TotalControl.
Even for a standard PC, you could buy a 16 port serial card and hook it up to 16 modems, either discreet devices or the dedicated ISP kit which might support dozens of incoming calls (possibly on a single bearer) via various means. Telebit netblazers and then ascend maxs were common in those days.
The question is if you could actually handle that many lines concurrently, with available software and the hardware resources of a single PC. Rachel's article as well as other comments here indicate that you could not.
From memory, it very much mattered on your choice of serial card/controller. High quality, high speed, (and high price) cards had their own buffers and better drivers resulting in fewer interrupts and more efficient use of host resources when supporting multiple concurrent file transfers at line speed. Being able to drive the modem-to-host connection faster than the modem connect speed was also helpful for modem protocols with compression.
You certainly could with later 386 and 486 systems. The multiport cards take care of the interrupt issues. 16 lines, all running at 115200 bps (to allow for better compression!), isn't even 2 megabits/sec. That is worst case. I had a 386 with ethernet and it could push 10 megabits no problem.
Telephone line modems were always slow compared to the microcomputers of their day. A 14400Baud modem will just send less than some 1.5KB/s. Even a Z80 could handle a few streams of that. I rather suspect this BBS offered fancier services than just chat and file up/download.
For sure. I knew people who ran multi-line BBS's on DOS PCs under DESQview, just like that (running Searchlight BBS, in my case). I know of a four line that was just using multiple external modems and non-standard IRQ's for COM3 and COM4 (since, by default, COM1/3 and COM2/4 share an IRQ).
MajorBBS could handle multiple lines on its own, but you had to handle ALL of the lines with one box. That meant a serial port interface like DigiBoard which provided some number (8 or 16 or more) of serial ports that you would connect to modems.
Yep - I ran a 16 line Major BBS back in the mid 90's in Seattle - used what was called a "Bocaboard" - had 16 serial ports on it - plugged in 16 external USR 28.8 modems. It all ran off of one PC.
I remember DigiBoard from my early ISP days. We attempted to turn a mid 90's-era Linux system (Slackware) into a terminal server. The Linux drivers for DigiBoard weren't quite up to it so we wound up going with Telebit Netblazers, I think.
Portmasters were very popular. Later on the ISP I worked with moved on to Ascend boxes which had digital modems (T1 / PRI lines.)
PRI was a huge step. The "individual modem" days were a mess. Each modem had a serial cable, phone line, and power brick. I remember doing some maintenance in one of the POPs. There were at least 100 modems, stacked on a cheap plastic shelving unit. The shelving unit was sagging from the weight and heat of all the modems.
This early POP was haphazardly built, so no cable management. I remember a river of phone cables coming out of the wall. The power bricks were also crazy. We had power strips 2 or 3 levels deep, making it a hazard to even get behind the rack without tripping on something.
They'd already switched to PRIs before I started so I missed out on that "fun", but I can personally vouch to the younguns here that every word you just wrote was completely plausible and likely.
You could just use DoubleDOS [1] to run the BBS in partition 0 and the door could run in parallel in partition 1. This was my setup before Windows 3 and DESQview replaced it.
DESQview never really clicked with me. Did run DoubleDOS and when I was a TA (or whatever it was) providing support in the computer lab in grad school (most people didn't have their own computers) it was a convenient way to switch between whatever official systems I needed to access and my own stuff.
My interpretation of the GP comment is that you are saying the same thing. Linux will return a pointer that is valid for your address space mappings, but might not be safe to actually use, because of VM overcommit. Unixes in general have no way to tell the process how much heap can be safely allocated.
All of these people who keep saying that webapps can replace desktop applications were simply never desktop power users. They don’t know what they don’t know.
reply