Messages in this thread | | | Date | Sun, 8 Aug 1999 17:50:01 -0400 (EDT) | From | "Mike A. Harris" <> | Subject | [OFFTOPIC] Re: First WinModem for Linux |
| |
On Sun, 8 Aug 1999, Riley Williams wrote:
> >>>> That's short-sighted, Mike. A Linmodem makes a perfectly fine > >>>> telco interface. It takes no more CPU than a buffered UART > >>>> running at 57600. > > >>>> See http:/linmodems.org for reasonable uses of a linmodem. > >There are TWO types of [lw]inmodems, one worth considering, one not.
Yes, but I personally don't consider either of them, for reasons that are related to compatibility mostly...
>Unfortunately, the latter is by far the commoner, and few people can >tell the difference, so the general advice to avoid them completely >still stands. The two types are: > > 1. [lw]inmodems with onboard DSP, where only the high level > firmware is implemented on the host PC. > > 2. [lw]inmodems without onboard DSP, where everything beyond > the telco isolation interface is implemented on the host > PC.
Right. Neither will work in minicom or Telix, and is one of my main concerns as well.
> >> Go back and read what he wrote until you understand why your > >> reply doesn't refute what Russ said. If you still don't > >> understand, check out the web site. (here's a hint - "modem" != > >> "telco interface") > >Here's another hint - type 2 [lw]inmodems are no more than telco >interfaces under a different name, and that is little more than a >marketing gimmick to allow the companies to sell their junk.
Exactly. ;o)
> > I don't care. As far as using a modem to connect to the > > internet, or to connect to another computer is concerned, a > > Winmodem uses and wastes CPU bandwidth and other valuable system > > resources. > >Nobody with any sense expects to get 100% CPU usage on any system with >a modem on it whilst the modem is in use, and that applies equally >well to true hardware modems as to [lw]inmodems of any type.
Well, any piece of hardware requires the CPU to do something to send it data. A real modem minimalizes this to the bare minimum. The higher the CPU to modem bps is set, the lower the overhead. I would consider a normal modem the baseline for comparison.
>However, people with type 2 [lw]inmodems will usually be lucky to get >more than 25% CPU usage on their systems because of the sheer load the >[lw]inmodem driver puts on their system - and, daft as it may sound, >the usable load they can put on their CPU's tends to go DOWN as the >CPU speed goes up, simply because the delays the drivers have to >introduce are fixed times, and the driver in general has to busy-wait >for the relevant interval.
Right.
>In comparison, type 1 [lw]inmodems will normally allow at least 70% >CPU usage for user applications, and hardware modems allow around 85% >to 90% CPU usage - and both those figures tend to go UP as the CPU >speed climbs.
I disagree. I am currently downloading RedHat hedwig updates from their site. I am using a REAL 33.6k modem, and am getting a constant rate of about 3.5 to 3.6k through the modem. I have no processes on the system consuming any CPU (all are idle). I just did a "top" reniced to 19 to watch the system load from the modem, etc..
Top consistently shows a 98.x% idle CPU, with the "user" portion fluctuating between 0.0 and 1.3 percent, average around .7 or so, and the system fluctuating between 1% and 5%. With the modem disabled, these numbers stay the same except system load is very slightly less. I would say negligable.
Hardware modems cause a CPU interrupt when the buffer is empty, and then it gets filled and the modem does it's thing. With even a 386 computer, a 56k modem doesn't load the system much at all. One could fill their system with UART's and have them all pumping data out fullstream, and not likely notice much going on at all. Even 8 kilobytes per second from a modem is nothing with a 40Mhz processor, let alone a 200Mhz processor.
So, I must be misunderstanding what you're meaning here, because I know that you know a lot about hardware, etc... Could you explain it a bit more what you mean?
> > The cost saved by using the winmodem, is wasted on the extra > > money put out for the host CPU. The unreliability of the things > > adds to the uselessness of them, as does their lack of > > portability. > >Whilst that is true of type 2 [lw]inmodems, it does not really apply >to type 1 [lw]inmodems
Well, I can't say for type 1 winmodems. I do know that aside from CPU and other resource beatings, that the things are NOT compatible with all software. Software that talks directly to the modem by opening the UART ports, will fail under any winmodem, and this is also a major cause of problem from a lot of people that I have worked with.
Compatibility is zero.
> > Nonetheless, manufacturers push them, and most people buy them > > unknowingly, because they do not know any better. Some people > > buy them because they are "winmodems" and that must mean that > > they are "better" in Windows. > >All too true, unfortunately...
Yep. Very unfortunately...
>Of those that cause problems, my experience says that over 90% will be >type 2 [lw]inmodems at that. In fact, daft as it sounds, the only non >type 2 [lw]inmodem I've come across that failed was a hardware modem >that had been subject to a lightning strike on the phone line it was >plugged into. I know several people with type 1 [lw]inmodems that have >never caused any problems, and several others with hardware modems >that have never caused any problems, but I only know one person with a >type 2 [lw]inmodem that doesn't cause him any problems, and that's >probably because he doesn't use it as he has an ethernet connection to >the internet - the modem came with the PC, and didn't cost him >anything...
Right. ;o) The things might work fine for someone that doesn't use them. ;o) Or someone that uses them minimally. Others get disconnected frequently, and never consider that it is their modem that causes this when they fire up 5 apps simultaneously for example. ;o)
> > Whats even funnier, is that even the people in the stores > > selling the crap, don't even know half the time. "Is this a > > winmodem", "no it is not", get home, put it in, Linux > > no-comprende, read manual - windows only.... back to store, "I > > said I DID NOT want a winmodem", "that isn't a winmodem", "Ok, I > > said I did not want a modem that ONLY WORKS IN WINDOWS NO MATTER > > WHAT IT'S CUTE LITTLE NAME IS", "oh, that modem should work..." > > Open manual, show them "windows only" - shocked look ... 'oh... > > um... ahh... try this modem instead..." > >That sounds like the staff at the local branch of PC World.
I've seen it happen in practically every computer store I've been in for a while. You overhear the salesperson selling the thing, and people blindly buy them up. The ones smart enough to ask, are dumb enough to trust the salesman's promise that it will work fine.
Even the boxes on the things are vague most of the time. I always state that the thing will be used in a Linux machine and that Winmodems and software modems of any shape or form as well as Plug and Play and PCI modems are completely unacceptable. The salesman looks confused right about then with a "well what else IS there? look on his face", and then when he comes back from talking to someone who does in fact know, I DEMAND that I get to RIP open the box and look the card over as well as the documentation. I would say a good 50% of the time it is STILL a winmodem/softmodem after all that. Imagine the poor soul that doesn't know the difference...
> > I live in an area where computer peripherals are not $5 for a > > winmodem, and $300 for a real modem like some other people imply > > they are. Here a crap software modem goes for about as low as > > $40-50 or so, and a real modem, with real hardware, goes for > > anywhere from $50-$200 or so. The $50 ones work just as good as > > the $200 ones from my experience, but they are REAL modems, and > > not software based. > >Locally, prices are as follows (all UK pounds): > > Hardware modem 30 to 350 > Type 1 20 to 50 > Type 2 15 to 70 > > ISDN 25 to 700
Your chart basically confirms my price statements. One could get a low end hardware modem for the same price or cheaper than one of the winmodem types except for the extreme low end which are likely to be replaced very soon after purchase. ;o)
>MIne is a USR Sportster 56k/VoiceFax hardware modem, and cost me all >of 40 pounds, and it's never given me any hassle at all.
USR's are over $100 here I believe. I put them in for customers with business systems, etc.. for their name brand. For other systems, I go with el-cheapo stuff which I've found to work admirably. Zoltrix for example: I've installed over 50 zoltrix modems (hardware modems) and not seen a single one go bad. They have 5 year warranties and are the cheapest of the lot to buy.
> > As such, ANYONE saving $10-$20 buy buying a winmodem, is getting > > taken. If someone purposefully chooses one, because it is > > cheaper after being warned what they are buying, then that is > > fine, their decision. > >I act as a consultancy locally, and my advice regarding modems is that >in general, what one saves buying a cheaper product, one will soon >spend on increased telephone bills.
Well, I don't see how that works unless you pay by the minute for local telephone access, but I generally agree. What you save on your winmodem, you will pay for replacing it down the road, maybe several times. Or, you will pay a consultant to come and install the drivers for it, or pay the computer store to get it working.
People often pay me my $30/hr to get their systems running. One of the problems is ALWAYS winmodems. Sometimes it takes me 10 minutes, sometimes 2 hours. At $30/hr, they're cheaper off getting a REAL modem - and I tell them that. They often go and buy it themselves, and come back with ANOTHER winmodem. ;o)
>For the manufacturers, the cost price difference between installing a >[lw]inmodem or a hardware modem is usually less than five pounds, and >well within their profit margins, so if one specifies that the system >be supplied with a hardware modem rather than a [lw]inmodem, they will >normally comply without any hassle.
Right, but only if it is stressed enough, and the people doing the installation are competant.
> > There might be a demand for them, if you wish to look at it like > > that, but I say the demand exists only because of the public's > > general lack of understanding of technology. The things are a > > burden to technology, and are a horrible thorn in the side of > > technology. > >IMHO, they are holding back technology, and the sooner they get booted >out of the market, the better. There's no way they can support ADSL >with them, and I firmly believe that ADSL technology would be much >more freely available were it not for the proliferation of >[lw]inmodems, and both the customers and the telco's would be much >better off as a result...
I completely agree as well. Winmodems are causing the hold back of high-speed internet to the masses.
> > Winmodems are a horrible disgrace to the technology age, as are > > all other software based crap hardware. Use any argument you > > like to counter, but my stance is very firm, and very well > > thought out, and covers many cases, based on my personal > > experience, and that of hundreds of RIPPED OFF customers. > >You'll get no arguments from me other than what I've expressed in this >missive, and my feeling is that we're pretty much in total agreement >on the subject.
I think that most people "in the know" so to speak, are also in agreement as well. I believe I misunderstood something that you said at the top though, so feel free to set me straight on it, or ignore it. We should move this to private email though, before the flames begin...
;o)
Take care Riley, TTYL
-- Mike A. Harris Linux advocate GNU advocate Computer Consultant Open Source advocate
Tea, Earl Grey, Hot...
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |