Re: [SOLVED!] Communicating with USB Modem
I was going to just reply to Phil privately, but there are a few points that could effect someone else's decision in a similar case, so I'm going to address them on list. Phil makes some good points, so I thought it appropriate to include my thoughts on them.
On Oct 18, 2010, at 12:38 AM, Phil Requirements wrote:
> On 2010-10-13 22:28:10 -0400, Hal Vaughan wrote:
>> Comments and some info at the bottom, so it makes sense when you read it. (Hey, there's NO way I'm going to top post on this list!)
>> Now I have a few notes on this, since there's been a few private
>> conversations with people who are in or have seen this thread.
>> 1) I'm basically making a "black box" for the people I'm working
>> with. I do NOT want them messing with the computer I give them.
>> Many times, in embedded systems, there is no PCI slot. Sometimes
>> there is, but I can't count on that. (I haven't picked my hardware
>> config yet.) So I really need an external modem since the only
>> connection I can count on is a USB connector. (Even with embedded,
>> most systems have 2 or more USB connectors, and some people may need
>> to hook up a printer there.)
> I see that here in Point 1 you rule out a PCI card-based solution
> while also admitting that the hardware has not yet been specified. It
> seems feasible that if you found a really great and cheap PCI modem,
> you could spec the hardware around that.
I spent close to a week researching possible mobos or all-in-one package systems. Some boards do have PCI slots, but some don't, so that's a big issue to consider. I'd much rather be able to spec the hardware around the mobo than a modem. Also, if I go for something like Soekris, which I like, but which may cost more than I want (they still haven't released final specs and cost on their Net6501, which is what I'd likely use), using a PCI card would be problematical. While I've seen PCI slots on their boards, I'd have to cut a hole in it to allow someone to plug a phone cable in.
That's one of the two biggest strikes against a PCI modem: I can't be sure how it'll fit in with the rest, and finding an appropriate mobo or mobo/case combo is a bigger concern than using a PCI modem. If I go in order of my reasoning, I listed that as strike 2 against PCI modems.
> I mention this because it seems to me that an external USB modem is
> not very "black box". There would be your discrete little appliance
> with a very conspicuous dongle coming off it. A PCI internal modem
> that fits inside the appliance would be much more black-box-like.
I agree. Until I read how you phrased it here, I had not thought of stating my reasoning here. I'll be working intelligent people who work in technical fields, but not in computer fields. The tradeoff is that if I need someone in another city to work with me so I can walk them through solving a problem so I don't have to make 4-10 hour round trip to get there and end up just flipping a switch, that's good, since they can work with me. But, on the flip side, when you're working with someone with an intelligent and technical mind, they're the ones more likely to think they can break the rules in some way because they know they're smart.
I'm more concerned about this being "black boxy" in terms of them not trying to log on or hook up a keyboard and monitor than anything else. I don't want them thinking of this box as a computer where, if it starts acting odd, they think they can work with.
On the other hand, even though I'm within 8 miles of the state capital building, I didn't get broadband for years, due to bureaucratic attitudes at the cable company (which kept changing ownership) and Verizon. I had to work with Linux and dial-up modems for several years. I also had to replace my PCI modems every few months. I don't remember why, but I do remember tossing out a lot of PCI modems that worked and stopped working. Sometimes it was due to a distro upgrade, other times it was a new modem that docs said would work that didn't.
If a PCI modem goes bad on me, then it means paying for them to ship me the unit overnight, me replacing it and sending it back overnight. If a USB modem goes bad, it means having a new one drop-shipped from Newegg. Considering package size, that's much easier and cheaper than replacing a PCI modem.
That, on my list, was strike 3 against PCI.
> The problem with PCI card modems is that they are not so likely to
> support GNU/Linux. So they aren't a clear winner. But they would be
> better for a black box.
And that was my strike 1 against PCI -- along with seeing some work now and not with later drivers.
As I said, my "black box" was a bit of a compromise. It's a black box as far as being something a user can log onto, but not 100% in that it has the external modem. Still, what I'm looking at now includes the embedded box, a PCI modem, and a wall-wart transformer. All the user has to do on receiving it is plug in a CAT5, plug in the USB modem, plug the RJ11 into the modem, and hook up and plug in the wall-wart. The modem does add 2 steps, but they're VERY simple steps.
>> 2) I can't count on having an RS232 interface, so that rules out a
>> lot of good modems.
> Another problem with these hardware modems is that they are large
> and require their own power supply, so they are not at all
> resembling a black box.
Yes. I use an RS232 on the system I'm using now, it's USR. It's about 1.5" high and 3"x7" base and needs its own wall-wart. While I like and use the flashing lights for trouble shooting, it also helps not have flashing lights on the unit the users will have.
>> 3) Yes, I know about US Robotics, I have known about them since my
>> Apple ][e days. (Yes, I'm that ancient!) However, their modem is
>> almost twice as much as this one and if things go well, I could need
>> 20 or more of these systems, and 20 * $20 = $400. That pays for 40
>> group Argentine tango lessons or almost pays for a 10 pack of
>> private ballroom lessons. I'd rather be dancing with women than
>> spending extra on modems when I don't have to. I know that's just
>> crazy and silly, but that's the way I am.
> Don't pay more than you have to, as long as everything works
> to your standard.
>> 4) I spent the better part of an afternoon and evening on trying to
>> get the Encore modem to work and couldn't, so I ordered the Rosewill
>> one. Yes, I'd love to have pursued it, but considering that, at
>> this point, I'm only working with one modem and not mass-ordering
>> them, it makes no sense to spend hours on making a modem work when
>> $30 will get me one that should work. Again, I know it's silly, but
>> I'd rather be dancing with women than working at my computer. At
>> some point, if an idea hits me, I may go back to working on the
>> Encore modem, but for now I don't see the point of investing more
>> time in it.
> I agree that for your purposes it does not make sense to try to make
> the Encore modem work, since you are trying to set up something worth
> replicating. Why spend many hours battling a modem if you don't have
And, as a Linux geek, there's the other factor: Often we get a piece of hardware and spend hours or, literally, days on making it work. When I started my business, I would spend 2 days to save $20 because I didn't have $20. I have income now and I would rather spend another $20 for a new modem that is documented as working than spend several days getting it to work. I have other goals and, as I've said, much more in life to do than be at the computer, so it was really a revelation for me when I realized, "This is a $15 part. Just get a new one. At this point, spending even 2 hours on it is a loss of time and money." True, as a geek, I wanted to "win," but there's no reason to, in this case. That won't help others who have the same modem, though.
>> Thanks to all that helped me with this. I would love it if I didn't
>> need to add another package, especially one not in repositories, but
>> this works and it doesn't take much to work, and including this
>> package as part of my setup won't be too hard at all. I don't know
>> what kind of reputation Rosewill has, but I've had good luck with
>> them, so I'll be sticking with their modem for now and when I get to
>> doing the mass-ordering as well.
>> If people still have more suggestions, I'm still open, but I'm not
>> going to spend hours Googling or researching or compiling to get the
>> Encore working.
>> Again, thanks for the help!
> Thank you for posting the update on this issue. It is interesting to
> know the outcome of all the work you are doing.
> I think you Rosewill solution is fine, though I would want to make
> sure it performs well over time before deploying two dozen of them.
Yes, I'll be testing it, but while I was waiting for it to come in, I started on another part of the program, so it'll be a while before I get back to the online routines and test them.
> I only responded because I have this nagging feeling that, as
> long as you're going to all this trouble, a PCI card-based solution
> would be a lot more black-box-like.
> On the other hand, I get the feeling that you do not want the R&D
> phase to go on for a long time, because it is cutting in to your other
> interests. In which case you might not want to investigate every
> possible approach.
I started what I'm doing in FEB of 2002 and worked solidly for years on it, until I was burned out and had to stop. Fortunately it was at the point that it was viable and the money kept coming in and the computers kept doing their jobs. I'm returning to it because I can raise my income, and the main reason for doing that is to fund the film production company I'm starting. So you're right. I want this work done because the sooner it's done, the sooner I can focus on my writing and film production. I like programming, but my passion is writing screenplays, so my rule for this phase is to save time in whatever way possible without compromising the important factors (like workability, usability, and security).
Thanks for your thoughts on this.