Red Hat has a history of publishing source, heeding free software licenses, and being nice to the developer community.
And, to be fair, SuSE hasn't been as well known for doing this. In particular, YaST (a major component of SuSE) has a commercial license, which leaves the rest of SuSE, basically, commercially licensed.
RedHat has a tendency to release what they've done under the GPL instead. (such as RPM) Heck, SuSE uses RPM, as well. In that way, RedHat has already helped SuSE out by releasing their work under the GPL.
RedHat could either go and start targetting advertising for people that are already using Linux and try and grab as much of the existing Linux market as they can, or they can try and target advertising towards people that aren't yet using Linux.
Traditionally, RedHat has done both. I've seen RedHat advertising in the Linux Journal as well as in other computer magazines (probably mostly ones aimed at Unix users or programmers, though).
I suspect that RedHat will continue to do both in Europe, too.
At the very least, any attempt to completely only target at potential new Linux users instead of existing Linux users would be suicidal if successful because part of what helps to get new Linux users using a particular distribution is that it's the distribution their long-time Linux using friend (or colleague or random people in a local LUG, whatever) either uses or recommends.
So, of course RedHat moving into Europe will take at least some business from SuSE, in the sense that there are people that might try SuSE and will instead try RedHat.
Hopefully they'll both manage to expand the general Linux market enough, however, that business will continue to expand for both of them.
From my reading (No, I'm not a lawyer) of that court case (Loudoun County Library) there wasn't anything prohibiting a school from some form of blocking.
Why? Because the court agreed that keeping children away from obscene materials was important. The problem with what they did was that they were blocking the free speech of *adults*. The final decision of the library board was to have optional filtering for adults and filtering for children that the parents could get turned off. And that's a policy that won't get challenged in court.
(And teachers and staff would, presumably, only have access via the school as part of their employement, and It'd be ridiculous to claim that there's anything wrong with an employer putting filters on employee net access)
They might try and do something with the other characters in the show (much like "Angel" spinning off from "Buffy The Vampire Slayer" this season.). Even if they tried to claim it was still the "X-Files", it would be a spin-off, however. Given that X-Files is so popular, it'd be pretty likely that they'd try to create some kind of spin-off.
If you're really interested in either increasing typing speed or decreasing typing injury (they're intrinsically linked) you should look into other kinds of keyboards.
My personal favorite (except for the cost, of course) is the DataHand which significantly decreases finger travel distance as well as pressure you need to apply. (those things will help you type faster.) It also lets you keep your wrists straight, which will help to reduce injury.
The DataHand is primarily designed to reduce injury, but once you learn it well, you'll be able to type faster; mostly due to shorter finger travel distance.
It's designed with an almost QWERTY layout, but you can always set it up as Dvorak and stick little notes onto the template (I think they'll even sell it to you with a Dvorak template.) The built in "mouse" is okay for cut-n-paste, but I'd suggest using it in addition to another pointing device (I prefer trackballs -- less arm movement required, even if they do suck for quake) (you can do this with GPM, under Linux) for when you're doing more mouse-intensive activities. (Like netscape)
If you're a computer programmer (like myself) and something like that could lengthen your career by one month, the cost is worth it. It's more likely to increase a career by, at least, several years.
Before the QWERTY keyboard layout it was straight alphabetic. QWERTY was specifically designed to space keys that were hit in combination frequently away from each other, so that a rapid typist wouldn't jam the arms as often as occured otherwise. This means that there was a tendency for common combinations to actually be on opposite hands (which is good for typing speed), but also a tendency for common letters not to be on the home row. For instance, 'e' is the single most commonly used letter in English. If I were to design a keyboard for rapidity of hitting keys, I'd put the 'e' key directly underneath a finger, not in a place one had to reach for it. "J" is fairly uncommon, but it's got one of the 4 best locations on the keyboard. (index and middle finger are strongest, and home row is shortest reach.)
The speed up for Dvorak over QWERTY isn't double, however, it's a small-ish percentage. It really isn't enough for an entire industry to switch (with all the costs of that). And, of course, momentum is hard to overcome. Nobody wants to build Dvorak keyboards if everybody knows QWERTY, and nobody wants to learn Dvorak if QWERTY keyboards are all that's available.
(The small-ish percentage increase in efficiency might be worth it if you get paid by the typed word or if you're developing RSI problems.)
Ellen Spertus wrote a paper titled "Why are There so Few Female Computer Scientists?". This was in 1991, but I don't think anything substantial has changed. (This, BTW, is a woman with a PhD in Electrical Engineering and Computer Science from MIT.)
I know the question is more about women in open-source projects than computer science, but people don't get involved, typically, with any kind of computer project (open source or otherwise) unless they've got the technical know-how to do so.
Anyways, the main points in her paper about why women don't get into computer science are:
Societal Factors. Such as women being discouraged from technical fields, mathematics, science and other "male" fields when they're children or in school.
Masculine Environment. That is, a lot of computer related workplaces or college classes have mostly men, and therefore a tendency for there to be things that some women would find offensive (sexist or sexual humour or female pin-ups, for example). Also, men use sarcasm or insults to communicate more often that women, often leaving women feeling as if the environment is hostile even when it isn't really. Different interests (sports, for example) can also leave women feeling less socially included.
Gender-biased language.
Some attempts to encourage women into these sorts of fields actually backfire, for instance making it seem as if women are less capable and that's why they might need extra help. (This can be subtle and the exact same thing could be interepreted differently by two different people.)
The paper goes into much more detail, however, including reasons why biology is probably not very much of a factor and some possible solutions.
If you read that paper I think you'll see that it relates fairly well to this topic.
I've run into those kinds of problems before, too. The problem in the cases I'd run into was that the two OS's didn't agree about the exact drive size and therefore didn't agree about how to partition it properly. If, say, one thinks you've got 1024 cylinders to deal with and the other thinks that you've got 1246 cylinders, the best strategy I found was to calculate sizes, determine how many cylinders that worked out to and leave a little bit of leeway (a small amount of unpartitioned space) between the two partitions, so that they didn't overwrite each other.
(yes, it's icky, it's strange, it sounds like black magic and it sounds similar to suggesting that you sacrafice a goat over your computer. But it's worked more than once. Check carefully that DOS or NT's fdisk and Linux's fdisk agree on number of cylinders. Or just use a boot disk to get into Linux.)
At one job, not all that long ago, I had somewhat similar requirements. The main difference, however, was that I needed to have nearly virgin installs of Win 3.1, Win 95, Win 98 and Win NT (4.0 Workstation) that I could recreate quickly. (The reason being that I needed to test the installation of a certain package and uninstalling all of the installed components and then trying an install again generated different results and was therefore not a useful test. Or at least not a test on a "like virgin" system anymore.)
Solution: two hard drives. Drive 0 (/dev/hda, C: in windows, whatever) was a 512MB drive (yes, nearly a throwaway, but all that was required for the OS and single piece of software) and Drive 1 (/dev/hdd, actually, IIRC -- slave on secondary with CD as master on secondary -- just to keep the two hard drives on separate IDE channels) was, I think, a 4 gig drive. You'll note that this is all very easy to scale up.
Anyways. Then I installed Linux (RedHat 6.0, minimal install) on the entire second drive (I don't even recall if I gave Linux any swap or not...) and created a boot floppy at the appropriate point.
Process:
Zero out first drive, "dd if=/dev/zero of=/dev/hda"
Create appropriate single FAT partition, format, etc.
Install appropriate OS (Win 3.1 was most difficult -- you'd be surprised how hard it was to scare up good installation media for MS-DOS 6)
Do anything with OS that would "render it typical" -- maybe play solitaire for five minutes, change colors around and reboot the OS a few times. (some of the OS's do something different on first boot or second boot from subsequent boots)
Stick in Linux boot floppy and reboot machine.
Copy entire first drive to a backup file: "dd if=/dev/hda of=win95-virgin-19980912.img" (or some filename appropriate like that for my purposes). Optionally gzip it (or gzip later). (which is why zero-ing it earlier can help.)
Repeat from step one with different OS.
Then, later, I would just "dd if=whatever.img of=/dev/hda", reboot and do whatever damage to that OS that I wanted, without having to worry about it. Then any states that I needed saved I could easily backup via Linux.
If you just use a Linux rescue disk and a *really* small Linux partition you could back up just the MBRs. The magic dd command line for that is like this: "dd if=/dev/hda of=MBR.img bs=512 count=1". For whatever reason, the LILO documentation reccomends something like "dd if=MBR.img of=/dev/hda bs=446 count=1" when you actually want to write the saved MBR back onto a drive.
More realistically, once you've installed each OS onto a separate partition and saved the MBR for it (do NT last), you can copy the MBR file onto the root of NT's boot drive (even if you have to do it via a floppy) and then: "attrib -s -r boot.ini" Edit boot.ini -- add a line like: C:\MBR95.IMG="Win95" Then be sure to "attrib +s +r boot.ini" afterwards. After that, NT should present you with the additional boot options when it starts up.
Look in the LILO documentation and "Linux+NT-Loader" mini-HOWTO for more info on that kind of strange trickery.
Some of the seeds that are now supplied to farmers have been modified to not be able to breed at all.
The motivation in that case, of course, isn't keeping genetic purity as much as keeping the farmers from taking some of the seed from a year's crop and just planting it next year, instead of buying more seed from the corporation.
More to the point: this kind of technique is getting to be reasonably advanced, I'm sure they could arrange so that the only plants outside of their sealed lab with that gene are incapable of actually breeding.
I would have thought that the Renunciates or "Amazons" in the Darkover series were more influental in redefining "female writing in the fantasy genre".
After all, the renunciates appeared much earlier in her writing, and it's the renunciates that you can find at least one book full of short stories written by fans of MZB and her renunciates.
(For those who haven't read her Darkover series: on Darkover there is a culture that is highly repressive of women, regarding them primarily as breeding material. However, there is a group of women called the "Renunciates" or sometimes "Amazons" who have renounced all their previous family ties in favor of allegiance to the rest of their group of women; they live outside the normal societal roles set for women.)
Perhaps Avalon sold more copies to the general (non-sci-fi fan or sci-fi author) community, though.
Also, I'd say that her writing was mostly more science-fiction than fantasy. Darkover and the psionic powers presented there (along with the psionic technology) is much closer to a science fiction kind of basis than true sword-and-sorcerer fantasy. Heck, the majority of the darkover series exists inside of the larger context of a galactic empire, which is hardly the stuff of fantasy novels. The main reason given in her books for the prevalence of swords instead of arrows or other missile weapons is that during Darkover's highest level of psionic technology the risk of destroying the entire planet with long-distance weapons caused everybody to agree not to use any form of long-distance weapon.
Mostly, this means I won't get to read any more of her excellent writing, since there won't be anymore. She truly was one of a fairly few really good authors out there that have been producing books for a long time. (It wasn't all that long ago that I read a book of hers written in 1968 -- it was still interesting and relevant, though the computers that appeared in it had the same problem as computers in ST:TOS -- big things with no visual interface.)
Geez, I think I'd be looking at other issues to see why your NetWare servers were crashing once a week. (NT, ok, maybe once a week I could understand)
I don't work there anymore, but it was a few years ago and the copy of NetWare was already out of date, the hardware was old for the time (I think it was a 386dx33 -- Pentium 90s were just coming onto the market when I quit), I only worked there one day a week and it was a small enough office that if the server happened to need to be rebooted once a week and be down for 5 minutes, it wasn't that big a deal... Besides, it usually happened during the backup after everybody left, anyways...
I'll take your word for it that NetWare is very stable -- I'm certain the NetWare instances I dealt with were all kinda screwed up, outdated and poorly administered. (I was never really the main guy in charge of any NetWare boxen -- just somebody available to help out sometimes. Linux boxen, a Solaris box and that one NT box, yes. (BTW, the NT box that crashed several times a day later got Linux installed on it, and after a few kernel upgrades and/or patches to get the right SMP stuff for the time it ran its 1.3.x Linux kernel for several months without crashing))
If there's more than one of something, it's no longer a single object. This applies also for points of failure.
"single point of failure" is the general term for this concept. It's not a perfect term, but it seems to be the most sane arrangement of the three important words. "Point of single failure" doesn't at all convey the right concept. That is, a single component or "point" in a system that can fail and cause the entire system to fail.
For instance, on a fileserver with a single SCSI card, a single SCSI drive and a single hard drive, you have 3 single points of failure. If any single one of those points fails, the entire server will fail, effectively. (that is, it will fail in its purpose of serving files to the network.) Another single point of failure would be a CPU -- and with Intel architecture, at least, two CPUs would give you two single points of failure.
When you're analyzing a system to determine (and improve) reliability, you determine all the single points of failure and attempt to eliminate them as single points of failure.
In the above example, if you had reliability concerns, I'd suggest replacing the single SCSI drive with a RAID array of some kind, either RAID 1 or RAID 5 -- that way one hard drive could fail and everything would still be fine. Then you'd no longer have a single point of failure in the hard drive because both drives would need to fail to cause failure from that. (and hard drives, in my experience, at least, are more prone to failure than SCSI cards or network cards.)
If you're really anal about reliability for a file server you'd have dual ethernet cards on two separate network segments, two SCSI RAID cards with RAID 5 arrays attached and a software RAID 1 array made from those (to survive either SCSI card dying), dual CPUs and an architecture and OS that could handle a CPU dying and then you'd clone the box and have a hot-swap backup machine with some kind of mirroring between them... (in other words, you'd be moving towards a cluster)
I believe the term you're looking for is "single point of failure".
In the novell environment you describe, the Novell server was still a single point of failure, but if it failed, there was a certain amount of redundancy elsewhere.
Another single point of failure would tend to be your incoming power -- the electricity goes out and all the machine go down. (except for the ones on a UPS that take a little while longer.) With a generator you could eliminate that single point of failure. You'd still have a number of single points of failure in the electrical wiring, but a chunk of copper only fails under extreme loads.
With the Sun and NC environment you have another single point of failure, the Sun server. Inside that Sun server you're likely to have some single points of failure, such as the OS itself, a few other bits of software, the hard disk, the hard disk controller, the network card, etc. (outside you'd have a hub or switch, too)
In my time I've had to deal with NetWare, Windows NT, Linux and Solaris (on an UltraSPARC) in some kind of administrative capacity. (mostly as the sysadmin) Under "interesting" load the NetWare box would crash maybe once a week and come back up fairly quickly (journaling file system). The NT box I had to deal with was a dual-PPro 200 and crashed several times a day. Since the NT box sometimes froze instead of rebooting, we eventually constructed a device that hooked up to the reset jumper of the NT box and the parallel port of the Linux box next to it and made it possible to hit the reset switch on the NT box via software run under Linux. The worst of the Linux boxes probably crashed once a month and came back up reasonably quickly (not as fast as the NetWare box, though). The only time the Solaris box crashed was due to crappy firmware in the Western Digital SCSI hard drives in an external RAID chassis that we got for it.
In other words, yes, you're right; if the server for those NCs goes down, they're all completely down. However, Solaris on SPARC architecture is, generally, really stable. It's pretty unlikely to crash more than once a school year. Heck, in the area I went to grade school the power went out more often than that.
Well, even if the kids are still able to get a shell and do a little bit of useful things that way, they'll probably be using a graphical interface (CDE, I'd bet) designed to hide the interesting Unix bits from them.
Letting your average kid loose on a Unix prompt would probably require a lot of hand-holding (that the teachers wouldn't necessarily be capable of). A nice clicky GUI interface with buttons for the most important things and menus for the slightly less important things would let those kids get stuff done. Sure, maybe a few kids could get a learning experience, but the idea at that age should be to teach them how to use a computer, in general, not anything too specific.
But the media is that way largely because that's what the average person wants.
All that violence is in the media because when it's in the media people seek it out. If one TV station suddenly stopped covering violence, except the actually important stuff, suddenly nobody would flip to that station to watch the news, their ratings would plummit, they'd get the message and start conforming again. (or go under)
Somebody helping somebody out? Sheesh, who wants to watch *that* on TV?
And, of course, who cares about the verity or verifiability of a news story anymore? Especially if a decent rumour that something might be true is so much more interesting?
You're thinking in terms of consumer modems. Most ISPs these days don't use normal consumer modems anymore.
Part of that, of course, is that it's a real bitch to manage that many individual modems, each with 3 cables running from it. (phone, serial, power) Modems that are built into a big rackmount chassis are easier, and even easier is a terminal server with the modems right in it. (especially PRI)
A second major reason for not using normal consumer modems is 56K. All of the "56K" (X2, V.90, K56flex) technologies require that the ISP end be "digitally terminated". IOW, that the ISP have an ISDN hookup instead of Plain-Old-Telephone analog service.
There are three common ways to get ISDN for an ISP. One is "BRI" which uses 2-pair for each BRI connection, but each BRI connection can handle two phone calls. (this is the kind of connection you'd get if you had ISDN at home.) You can also get a T1 line provisioned for ISDN, usually called a "PRI". (Primary Rate Interface, IIRC) A T1 (in the US) can handle 24 regular phone calls. When provisioned as a PRI you get 23 ISDN B-channel phone lines (the 24th is a control channel). (I understand than in Europe they get 30 lines per PRI)
If you were to get, say, a Livingston^H^H^H^H^H^H^H^H^H^HLucent PM3 you could stick modem cards into it, use 4 (or maybe 3 or 5...) rackmount units, have a single ethernet connection, a single power connection and two PRI lines (size of ethernet, so not much hassle) and handle 46 calls. That's much easier to manage than 46 individual modems and the terminal server(s) to go with it.
Now, at one point in time I worked at an ISP that had a few different kinds of equipment capable of handling 56K. Starting with USR "Total Control" 16 BRI modems per rackmount talking to terminal servers to PM3s. Some of this equipment supported caller-id, and on the exact same kind of phone line as some stuff that didn't. For the ones that did, our RADIUS logs included the phone number somebody dialed in from when that information was available.
So, that's the first way you're wrong -- for an ISP to support caller-id is trivial and is just a matter of purchasing the right equipment or possibly even just a matter of using the right soft(firm)ware for your PRI terminal servers.
The second way you're wrong is that he's talking about a special service over PCS, not a normal modem connection. It's probably got some kind of software on their end that is, basically, a proxy. Even if it really was more like a conventional modem/terminal server setup, transparent proxies are out there and not all that hard to do, even.
However, you've got a point about the potential marketing usage of the information -- Also, calling somebody on their cell phone for a marketing call would probably be illegal for the same reason faxing unsolicited advertising is illegal. (costs the recipient to receive and they don't get a choice)
If there's one group that I trust to honor privacy even less than any of our governments it's large corporations. And if there's any group that I trust less than large corporations to honor privacy, it's crackers.
Does anybody know if these crackers are anything more than greedy script kiddies?
With this kind of thing (governments eroding privacy, eroding any attempts to use encryption, private sector being even worse about privacy, etc.) the average law-abiding citizen in any country might as well post a daily log of all their activities and financial statements to USENET, because everybody could get to the info anyways.
And, jeez, how hard is it, really, to separate a bank network from the internet entirely and only allow absolutely necessary things through firewalls? (and to keep computers up-to-date, for that matter) Or is this all mostly being done by people that manage to get access (somehow) to terminals in the banks themselves?
In order for Brin's vision to occur, you need technology that isn't here quite yet, but could probably happen in the not-too-distant future.
Specifically: cheap miniature cameras combined with some kind of computer and broadcast capability. If you could make all that small enough to fit into a pair of sunglasses (perhaps with variable darkness of the glasses as well or maybe a display so you could "zoom" in on things, etc.) then the average citizen could record and/or broadcast *everything* that happened to them.
Combine that with cameras mounted in public places that broadcast their images to whomever wants it.
Then add in easier access to other surveillance technology, such as infrared, that radar that can see through walls, high resolution satellite cameras, etc.
Since the rich and powerful are more interesting, they're the ones that would end up with various cameras (or whatever) that would make it possible to monitor everywhere that they go as well as their general location inside of a house, etc. That means that it would be very difficult for the rich and powerful to meet somebody to talk with them secretly.
I'm still not convinced that this would be a good future, though. The ultimate in inhumane prisons is the one where somebody can always see what you're doing. Without *any* real privacy, we'd all eventually go mad. (though, if you "went postal", everybody would know that you'd bought a gun and were carrying it with you right then...)
However, the GPL specifically states "Activities other than copying, distribution and modification are not covered by this License; they are outside its scope.".
And, more relevantly: "3. You may copy and distribute the Program (or a work based on it, under Section 2) in object code or executable form under the terms of Sections 1 and 2 above provided that you also do one of the following:
a) Accompany it with the complete corresponding machine-readable source code, which must be distributed under the terms of Sections 1 and 2 above on a medium customarily used for software interchange; or,
b) Accompany it with a written offer, valid for at least three years, to give any third party, for a charge no more than your cost of physically performing source distribution, a complete machine-readable copy of the corresponding source code, to be distributed under the terms of Sections 1 and 2 above on a medium customarily used for software interchange; or," [possiblity "c" omitted, as it doesn't apply unless you got things under condition "b" in the first place.]
Now, the next section is also even more relevant: "4. You may not copy, modify, sublicense, or distribute the Program except as expressly provided under this License. Any attempt otherwise to copy, modify, sublicense or distribute the Program is void, and will automatically terminate your rights under this License. However, parties who have received copies, or rights, from you under this License will not have their licenses terminated so long as such parties remain in full compliance."
And, of course: " 6. Each time you redistribute the Program (or any work based on the Program), the recipient automatically receives a license from the original licensor to copy, distribute or modify the Program subject to these terms and conditions. You may not impose any further restrictions on the recipients' exercise of the rights granted herein. You are not responsible for enforcing compliance by third parties to this License. "
IOW, Any distribution of software under the GPL license must be accompanied by source code or an offer of the source code. Also, GPL'd software can't be placed under another license.
There's nothing in there that puts any restrictions on what you can do with the software all on your own, without sharing it with any friends. That doesn't mean that it's not under the GPL, just that the GPL doesn't apply.
Note, also, that if you were to download and modify a copy of a GPL program, modify it and then share it with one friend, you'd only be obligated by the license to share the source code with that one friend. If they then go and share it with a fourth party (1st party being original developer, you as second party, your friend is third party) you're not suddenly obligated to share the source code with that fourth party -- your friend is, however. (IOW, the distribution requirements don't work virally backwards up the chain)
I think Corel should, instead, have released it under the GPL with a strongly worded request that people not redistribute that Beta version and instead wait for the final version to come out. That way they wouldn't conflict with the GPL and they'd still manage to, largely, keep it from being distributed too widely.
Beta testing isn't "releasing" a product. I'd have to say that it does count as "distributing" a product, however, and that's what the GPL addresses, not "releases".
(note that SuSE has currently dropped it from their distribution.)
Personally, given that it's had more than one different security problem pop up in a relatively short time, I'd wait till they fix the existing bugs and then give it another month to see if any other bugs pop up before using it. (But, then, my need to worry at all about any kind of FTP service dissappeared last week and is unlikely to return.)
With credit card processing there is a bank network that you need to connect to somehow. The most traditional method is a modem built into a credit card terminal (what you see next to the cash register in brick-and-mortar stores).
Obviously, actually typing in credit card numbers manually is way too much work for an online store, but you still have to connect to that network somehow to do the processing.
Method #1: A modem hooked up to the computer. This is that CCVS and OpenCCVS like to do. (You could also do this with a leased line, but you'd need to be processing a *lot* of transactions to justify a leased line.)
Method #2: Connect over the internet to somebody else that does the connection to the "clearing house" for you. This is what CyberCash does. (CyberCash also has their own "cash" that customers can use, which is much better for small amounts of cash per purchase.)
I've dealt with CyberCash to some degree or another at two different jobs. Once you've got it set up, it's okay, as long as you don't have large volumes to work with. Cybercash has two basic problems. First off, it does everything over the internet, so it is at most as reliable as the internet. Secondly, it uses HTTP over TCP to perform the transactions. What this all means is that any kind of problems that occur will tend to leave you unsure whether or not the transaction has actually gone through or not. With older versions there was a database on your end that kept track of transactions that could get out of sync (with the fixes involving direct modification of the database) with their version of things. Newer versions of their protocol have "fixed" this problem by moving the database to their end. So if you deal with maybe one charge an hour (or 10 an hour, whatever), CyberCash will generally seem okay. When you have to deal with batches of about a thousand charges once a month, the problems start to become intolerable.
My suggestion: investigate the various options that involve your computer dialing the clearing house itself. Make *certain* that the software is capable of using a queue in an intelligent fashion (ie, if it has two charges waiting, dialing once to clear both), make sure the software has *some* way to handle more than one modem (if it comes to that) and try hard to find a bank that you can use ISDN with. (Since ISDN connects almost instantly)
(Note: I think Cybercash does have one competitor, but I don't think they support Linux yet and I suspect they'll have the same basic unreliability of the internet problems that Cybercash does)
(Note: I've basically done this before, but only from RH4.2 to RH5.2 -- it's probably changed little in RH6))
Create a customized RedHat installation.
Mirror the RedHat distribution.
copy the RPMs you want into the big RPMS directory, remove RPMS you don't want, etc. I'd highly reccomend replacing old RPMs with the updates and grabbing the ssh (etc) rpms from ftp.replay.com. Also, install autorpm and either customize it or make a separate RPM with just the files for autorpm to use an updates mirror that you make.
Edit RedHat/base/comps -- the format is somewhat self-explanatory. For the purpose you're describing, go to the "0 --hide Workstation" section and customize that. It's probably easiest to make that list by doing a completely custom install off of the mirror you made in step one and then doing rpm -qa --queryformat '%{NAME}\n' to get a list of all the packages that got installed.
run genhdlist on the appropriate directory.
Put that all up someplace
Optionally, do the whole kickstart thing.
Set up an updates "mirror" someplace and point autorpm on all those workstations at that mirror. Presumably set up to only install from your relatively secure mirror and to check PGP signatures, etc...
You can put your custom install on an NFS mount point or burn it onto a CD (possibly even a bootable CD) and install all those machines... If you had more than 20 machines I'd suggest using DHCP, but you can probably manage without with 20 machines. You still might want to consider using DHCP to assign addresses to minimize how much effort you need to go through when you do the installs.
Slashdot Mirror
Red Hat has a history of publishing source, heeding free software licenses, and being nice to the developer community.
And, to be fair, SuSE hasn't been as well known for doing this. In particular, YaST (a major component of SuSE) has a commercial license, which leaves the rest of SuSE, basically, commercially licensed.
RedHat has a tendency to release what they've done under the GPL instead. (such as RPM) Heck, SuSE uses RPM, as well. In that way, RedHat has already helped SuSE out by releasing their work under the GPL.
RedHat could either go and start targetting advertising for people that are already using Linux and try and grab as much of the existing Linux market as they can, or they can try and target advertising towards people that aren't yet using Linux.
Traditionally, RedHat has done both. I've seen RedHat advertising in the Linux Journal as well as in other computer magazines (probably mostly ones aimed at Unix users or programmers, though).
I suspect that RedHat will continue to do both in Europe, too.
At the very least, any attempt to completely only target at potential new Linux users instead of existing Linux users would be suicidal if successful because part of what helps to get new Linux users using a particular distribution is that it's the distribution their long-time Linux using friend (or colleague or random people in a local LUG, whatever) either uses or recommends.
So, of course RedHat moving into Europe will take at least some business from SuSE, in the sense that there are people that might try SuSE and will instead try RedHat.
Hopefully they'll both manage to expand the general Linux market enough, however, that business will continue to expand for both of them.
From my reading (No, I'm not a lawyer) of that court case (Loudoun County Library) there wasn't anything prohibiting a school from some form of blocking.
Why? Because the court agreed that keeping children away from obscene materials was important. The problem with what they did was that they were blocking the free speech of *adults*. The final decision of the library board was to have optional filtering for adults and filtering for children that the parents could get turned off. And that's a policy that won't get challenged in court.
(And teachers and staff would, presumably, only have access via the school as part of their employement, and It'd be ridiculous to claim that there's anything wrong with an employer putting filters on employee net access)
They might try and do something with the other characters in the show (much like "Angel" spinning off from "Buffy The Vampire Slayer" this season.). Even if they tried to claim it was still the "X-Files", it would be a spin-off, however. Given that X-Files is so popular, it'd be pretty likely that they'd try to create some kind of spin-off.
If you're really interested in either increasing typing speed or decreasing typing injury (they're intrinsically linked) you should look into other kinds of keyboards.
My personal favorite (except for the cost, of course) is the DataHand which significantly decreases finger travel distance as well as pressure you need to apply. (those things will help you type faster.) It also lets you keep your wrists straight, which will help to reduce injury.
The DataHand is primarily designed to reduce injury, but once you learn it well, you'll be able to type faster; mostly due to shorter finger travel distance.
It's designed with an almost QWERTY layout, but you can always set it up as Dvorak and stick little notes onto the template (I think they'll even sell it to you with a Dvorak template.) The built in "mouse" is okay for cut-n-paste, but I'd suggest using it in addition to another pointing device (I prefer trackballs -- less arm movement required, even if they do suck for quake) (you can do this with GPM, under Linux) for when you're doing more mouse-intensive activities. (Like netscape)
If you're a computer programmer (like myself) and something like that could lengthen your career by one month, the cost is worth it. It's more likely to increase a career by, at least, several years.
Before the QWERTY keyboard layout it was straight alphabetic. QWERTY was specifically designed to space keys that were hit in combination frequently away from each other, so that a rapid typist wouldn't jam the arms as often as occured otherwise. This means that there was a tendency for common combinations to actually be on opposite hands (which is good for typing speed), but also a tendency for common letters not to be on the home row. For instance, 'e' is the single most commonly used letter in English. If I were to design a keyboard for rapidity of hitting keys, I'd put the 'e' key directly underneath a finger, not in a place one had to reach for it. "J" is fairly uncommon, but it's got one of the 4 best locations on the keyboard. (index and middle finger are strongest, and home row is shortest reach.)
The speed up for Dvorak over QWERTY isn't double, however, it's a small-ish percentage. It really isn't enough for an entire industry to switch (with all the costs of that). And, of course, momentum is hard to overcome. Nobody wants to build Dvorak keyboards if everybody knows QWERTY, and nobody wants to learn Dvorak if QWERTY keyboards are all that's available.
(The small-ish percentage increase in efficiency might be worth it if you get paid by the typed word or if you're developing RSI problems.)
Ellen Spertus wrote a paper titled "Why are There so Few Female Computer Scientists?". This was in 1991, but I don't think anything substantial has changed. (This, BTW, is a woman with a PhD in Electrical Engineering and Computer Science from MIT.)
I know the question is more about women in open-source projects than computer science, but people don't get involved, typically, with any kind of computer project (open source or otherwise) unless they've got the technical know-how to do so.
Anyways, the main points in her paper about why women don't get into computer science are:
- Societal Factors. Such as women being discouraged from technical fields, mathematics, science and other "male" fields when they're children or in school.
- Masculine Environment. That is, a lot of computer related workplaces or college classes have mostly men, and therefore a tendency for there to be things that some women would find offensive (sexist or sexual humour or female pin-ups, for example). Also, men use sarcasm or insults to communicate more often that women, often leaving women feeling as if the environment is hostile even when it isn't really. Different interests (sports, for example) can also leave women feeling less socially included.
- Gender-biased language.
- Some attempts to encourage women into these sorts of fields actually backfire, for instance making it seem as if women are less capable and that's why they might need extra help. (This can be subtle and the exact same thing could be interepreted differently by two different people.)
The paper goes into much more detail, however, including reasons why biology is probably not very much of a factor and some possible solutions.If you read that paper I think you'll see that it relates fairly well to this topic.
I've run into those kinds of problems before, too. The problem in the cases I'd run into was that the two OS's didn't agree about the exact drive size and therefore didn't agree about how to partition it properly. If, say, one thinks you've got 1024 cylinders to deal with and the other thinks that you've got 1246 cylinders, the best strategy I found was to calculate sizes, determine how many cylinders that worked out to and leave a little bit of leeway (a small amount of unpartitioned space) between the two partitions, so that they didn't overwrite each other.
(yes, it's icky, it's strange, it sounds like black magic and it sounds similar to suggesting that you sacrafice a goat over your computer. But it's worked more than once. Check carefully that DOS or NT's fdisk and Linux's fdisk agree on number of cylinders. Or just use a boot disk to get into Linux.)
Solution: two hard drives. Drive 0 (/dev/hda, C: in windows, whatever) was a 512MB drive (yes, nearly a throwaway, but all that was required for the OS and single piece of software) and Drive 1 (/dev/hdd, actually, IIRC -- slave on secondary with CD as master on secondary -- just to keep the two hard drives on separate IDE channels) was, I think, a 4 gig drive. You'll note that this is all very easy to scale up.
Anyways. Then I installed Linux (RedHat 6.0, minimal install) on the entire second drive (I don't even recall if I gave Linux any swap or not...) and created a boot floppy at the appropriate point.
Process:
Then, later, I would just "dd if=whatever.img of=/dev/hda", reboot and do whatever damage to that OS that I wanted, without having to worry about it. Then any states that I needed saved I could easily backup via Linux.
If you just use a Linux rescue disk and a *really* small Linux partition you could back up just the MBRs. The magic dd command line for that is like this: "dd if=/dev/hda of=MBR.img bs=512 count=1". For whatever reason, the LILO documentation reccomends something like "dd if=MBR.img of=/dev/hda bs=446 count=1" when you actually want to write the saved MBR back onto a drive.
More realistically, once you've installed each OS onto a separate partition and saved the MBR for it (do NT last), you can copy the MBR file onto the root of NT's boot drive (even if you have to do it via a floppy) and then:
"attrib -s -r boot.ini"
Edit boot.ini -- add a line like:
C:\MBR95.IMG="Win95"
Then be sure to "attrib +s +r boot.ini" afterwards. After that, NT should present you with the additional boot options when it starts up.
Look in the LILO documentation and "Linux+NT-Loader" mini-HOWTO for more info on that kind of strange trickery.
Some of the seeds that are now supplied to farmers have been modified to not be able to breed at all.
The motivation in that case, of course, isn't keeping genetic purity as much as keeping the farmers from taking some of the seed from a year's crop and just planting it next year, instead of buying more seed from the corporation.
More to the point: this kind of technique is getting to be reasonably advanced, I'm sure they could arrange so that the only plants outside of their sealed lab with that gene are incapable of actually breeding.
I would have thought that the Renunciates or "Amazons" in the Darkover series were more influental in redefining "female writing in the fantasy genre".
After all, the renunciates appeared much earlier in her writing, and it's the renunciates that you can find at least one book full of short stories written by fans of MZB and her renunciates.
(For those who haven't read her Darkover series: on Darkover there is a culture that is highly repressive of women, regarding them primarily as breeding material. However, there is a group of women called the "Renunciates" or sometimes "Amazons" who have renounced all their previous family ties in favor of allegiance to the rest of their group of women; they live outside the normal societal roles set for women.)
Perhaps Avalon sold more copies to the general (non-sci-fi fan or sci-fi author) community, though.
Also, I'd say that her writing was mostly more science-fiction than fantasy. Darkover and the psionic powers presented there (along with the psionic technology) is much closer to a science fiction kind of basis than true sword-and-sorcerer fantasy. Heck, the majority of the darkover series exists inside of the larger context of a galactic empire, which is hardly the stuff of fantasy novels. The main reason given in her books for the prevalence of swords instead of arrows or other missile weapons is that during Darkover's highest level of psionic technology the risk of destroying the entire planet with long-distance weapons caused everybody to agree not to use any form of long-distance weapon.
Mostly, this means I won't get to read any more of her excellent writing, since there won't be anymore. She truly was one of a fairly few really good authors out there that have been producing books for a long time. (It wasn't all that long ago that I read a book of hers written in 1968 -- it was still interesting and relevant, though the computers that appeared in it had the same problem as computers in ST:TOS -- big things with no visual interface.)
Geez, I think I'd be looking at other issues to see why your NetWare servers were crashing once a week. (NT, ok, maybe once a week I could understand)
I don't work there anymore, but it was a few years ago and the copy of NetWare was already out of date, the hardware was old for the time (I think it was a 386dx33 -- Pentium 90s were just coming onto the market when I quit), I only worked there one day a week and it was a small enough office that if the server happened to need to be rebooted once a week and be down for 5 minutes, it wasn't that big a deal... Besides, it usually happened during the backup after everybody left, anyways...
I'll take your word for it that NetWare is very stable -- I'm certain the NetWare instances I dealt with were all kinda screwed up, outdated and poorly administered. (I was never really the main guy in charge of any NetWare boxen -- just somebody available to help out sometimes. Linux boxen, a Solaris box and that one NT box, yes. (BTW, the NT box that crashed several times a day later got Linux installed on it, and after a few kernel upgrades and/or patches to get the right SMP stuff for the time it ran its 1.3.x Linux kernel for several months without crashing))
If there's more than one of something, it's no longer a single object. This applies also for points of failure.
"single point of failure" is the general term for this concept. It's not a perfect term, but it seems to be the most sane arrangement of the three important words. "Point of single failure" doesn't at all convey the right concept. That is, a single component or "point" in a system that can fail and cause the entire system to fail.
For instance, on a fileserver with a single SCSI card, a single SCSI drive and a single hard drive, you have 3 single points of failure. If any single one of those points fails, the entire server will fail, effectively. (that is, it will fail in its purpose of serving files to the network.) Another single point of failure would be a CPU -- and with Intel architecture, at least, two CPUs would give you two single points of failure.
When you're analyzing a system to determine (and improve) reliability, you determine all the single points of failure and attempt to eliminate them as single points of failure.
In the above example, if you had reliability concerns, I'd suggest replacing the single SCSI drive with a RAID array of some kind, either RAID 1 or RAID 5 -- that way one hard drive could fail and everything would still be fine. Then you'd no longer have a single point of failure in the hard drive because both drives would need to fail to cause failure from that. (and hard drives, in my experience, at least, are more prone to failure than SCSI cards or network cards.)
If you're really anal about reliability for a file server you'd have dual ethernet cards on two separate network segments, two SCSI RAID cards with RAID 5 arrays attached and a software RAID 1 array made from those (to survive either SCSI card dying), dual CPUs and an architecture and OS that could handle a CPU dying and then you'd clone the box and have a hot-swap backup machine with some kind of mirroring between them... (in other words, you'd be moving towards a cluster)
I believe the term you're looking for is "single point of failure".
In the novell environment you describe, the Novell server was still a single point of failure, but if it failed, there was a certain amount of redundancy elsewhere.
Another single point of failure would tend to be your incoming power -- the electricity goes out and all the machine go down. (except for the ones on a UPS that take a little while longer.) With a generator you could eliminate that single point of failure. You'd still have a number of single points of failure in the electrical wiring, but a chunk of copper only fails under extreme loads.
With the Sun and NC environment you have another single point of failure, the Sun server. Inside that Sun server you're likely to have some single points of failure, such as the OS itself, a few other bits of software, the hard disk, the hard disk controller, the network card, etc. (outside you'd have a hub or switch, too)
In my time I've had to deal with NetWare, Windows NT, Linux and Solaris (on an UltraSPARC) in some kind of administrative capacity. (mostly as the sysadmin) Under "interesting" load the NetWare box would crash maybe once a week and come back up fairly quickly (journaling file system). The NT box I had to deal with was a dual-PPro 200 and crashed several times a day. Since the NT box sometimes froze instead of rebooting, we eventually constructed a device that hooked up to the reset jumper of the NT box and the parallel port of the Linux box next to it and made it possible to hit the reset switch on the NT box via software run under Linux. The worst of the Linux boxes probably crashed once a month and came back up reasonably quickly (not as fast as the NetWare box, though). The only time the Solaris box crashed was due to crappy firmware in the Western Digital SCSI hard drives in an external RAID chassis that we got for it.
In other words, yes, you're right; if the server for those NCs goes down, they're all completely down. However, Solaris on SPARC architecture is, generally, really stable. It's pretty unlikely to crash more than once a school year. Heck, in the area I went to grade school the power went out more often than that.
Well, even if the kids are still able to get a shell and do a little bit of useful things that way, they'll probably be using a graphical interface (CDE, I'd bet) designed to hide the interesting Unix bits from them.
Letting your average kid loose on a Unix prompt would probably require a lot of hand-holding (that the teachers wouldn't necessarily be capable of). A nice clicky GUI interface with buttons for the most important things and menus for the slightly less important things would let those kids get stuff done. Sure, maybe a few kids could get a learning experience, but the idea at that age should be to teach them how to use a computer, in general, not anything too specific.
They're probably taking credit for Digital's help to the open source community.
But, then, it seems like Digital has been helping out since sometime before 1994... Oh well.
But the media is that way largely because that's what the average person wants.
All that violence is in the media because when it's in the media people seek it out. If one TV station suddenly stopped covering violence, except the actually important stuff, suddenly nobody would flip to that station to watch the news, their ratings would plummit, they'd get the message and start conforming again. (or go under)
Somebody helping somebody out? Sheesh, who wants to watch *that* on TV?
And, of course, who cares about the verity or verifiability of a news story anymore? Especially if a decent rumour that something might be true is so much more interesting?
You're thinking in terms of consumer modems. Most ISPs these days don't use normal consumer modems anymore.
Part of that, of course, is that it's a real bitch to manage that many individual modems, each with 3 cables running from it. (phone, serial, power) Modems that are built into a big rackmount chassis are easier, and even easier is a terminal server with the modems right in it. (especially PRI)
A second major reason for not using normal consumer modems is 56K. All of the "56K" (X2, V.90, K56flex) technologies require that the ISP end be "digitally terminated". IOW, that the ISP have an ISDN hookup instead of Plain-Old-Telephone analog service.
There are three common ways to get ISDN for an ISP. One is "BRI" which uses 2-pair for each BRI connection, but each BRI connection can handle two phone calls. (this is the kind of connection you'd get if you had ISDN at home.) You can also get a T1 line provisioned for ISDN, usually called a "PRI". (Primary Rate Interface, IIRC) A T1 (in the US) can handle 24 regular phone calls. When provisioned as a PRI you get 23 ISDN B-channel phone lines (the 24th is a control channel). (I understand than in Europe they get 30 lines per PRI)
If you were to get, say, a Livingston^H^H^H^H^H^H^H^H^H^HLucent PM3 you could stick modem cards into it, use 4 (or maybe 3 or 5...) rackmount units, have a single ethernet connection, a single power connection and two PRI lines (size of ethernet, so not much hassle) and handle 46 calls. That's much easier to manage than 46 individual modems and the terminal server(s) to go with it.
Now, at one point in time I worked at an ISP that had a few different kinds of equipment capable of handling 56K. Starting with USR "Total Control" 16 BRI modems per rackmount talking to terminal servers to PM3s. Some of this equipment supported caller-id, and on the exact same kind of phone line as some stuff that didn't. For the ones that did, our RADIUS logs included the phone number somebody dialed in from when that information was available.
So, that's the first way you're wrong -- for an ISP to support caller-id is trivial and is just a matter of purchasing the right equipment or possibly even just a matter of using the right soft(firm)ware for your PRI terminal servers.
The second way you're wrong is that he's talking about a special service over PCS, not a normal modem connection. It's probably got some kind of software on their end that is, basically, a proxy. Even if it really was more like a conventional modem/terminal server setup, transparent proxies are out there and not all that hard to do, even.
However, you've got a point about the potential marketing usage of the information -- Also, calling somebody on their cell phone for a marketing call would probably be illegal for the same reason faxing unsolicited advertising is illegal. (costs the recipient to receive and they don't get a choice)
If there's one group that I trust to honor privacy even less than any of our governments it's large corporations. And if there's any group that I trust less than large corporations to honor privacy, it's crackers.
Does anybody know if these crackers are anything more than greedy script kiddies?
With this kind of thing (governments eroding privacy, eroding any attempts to use encryption, private sector being even worse about privacy, etc.) the average law-abiding citizen in any country might as well post a daily log of all their activities and financial statements to USENET, because everybody could get to the info anyways.
And, jeez, how hard is it, really, to separate a bank network from the internet entirely and only allow absolutely necessary things through firewalls? (and to keep computers up-to-date, for that matter) Or is this all mostly being done by people that manage to get access (somehow) to terminals in the banks themselves?
In order for Brin's vision to occur, you need technology that isn't here quite yet, but could probably happen in the not-too-distant future.
Specifically: cheap miniature cameras combined with some kind of computer and broadcast capability. If you could make all that small enough to fit into a pair of sunglasses (perhaps with variable darkness of the glasses as well or maybe a display so you could "zoom" in on things, etc.) then the average citizen could record and/or broadcast *everything* that happened to them.
Combine that with cameras mounted in public places that broadcast their images to whomever wants it.
Then add in easier access to other surveillance technology, such as infrared, that radar that can see through walls, high resolution satellite cameras, etc.
Since the rich and powerful are more interesting, they're the ones that would end up with various cameras (or whatever) that would make it possible to monitor everywhere that they go as well as their general location inside of a house, etc. That means that it would be very difficult for the rich and powerful to meet somebody to talk with them secretly.
I'm still not convinced that this would be a good future, though. The ultimate in inhumane prisons is the one where somebody can always see what you're doing. Without *any* real privacy, we'd all eventually go mad. (though, if you "went postal", everybody would know that you'd bought a gun and were carrying it with you right then...)
The GPL does follow a product at all points.
However, the GPL specifically states "Activities other than copying, distribution and modification are not covered by this License; they are outside its scope.".
And, more relevantly:
"3. You may copy and distribute the Program (or a work based on it, under Section 2) in object code or executable form under the terms of Sections 1 and 2 above provided that you also do one of the following:
a) Accompany it with the complete corresponding machine-readable source code, which must be distributed under the terms of Sections 1 and 2 above on a medium customarily used for software interchange; or,
b) Accompany it with a written offer, valid for at least three years, to give any third party, for a charge no more than your cost of physically performing source distribution, a complete machine-readable copy of the corresponding source code, to be distributed under the terms of Sections 1 and 2 above on a medium customarily used for software interchange; or,"
[possiblity "c" omitted, as it doesn't apply unless you got things under condition "b" in the first place.]
Now, the next section is also even more relevant:
"4. You may not copy, modify, sublicense, or distribute the Program except as expressly provided under this License. Any attempt otherwise to copy, modify, sublicense or distribute the Program is void, and will automatically terminate your rights under this License. However, parties who have received copies, or rights, from you under this License will not have their licenses terminated so long as such parties remain in full compliance."
And, of course:
"
6. Each time you redistribute the Program (or any work based on the Program), the recipient automatically receives a license from the original licensor to copy, distribute or modify the Program subject to these terms and conditions. You may not impose any further restrictions on the recipients' exercise of the rights granted herein. You are not responsible for enforcing compliance by third parties to this License.
"
IOW, Any distribution of software under the GPL license must be accompanied by source code or an offer of the source code. Also, GPL'd software can't be placed under another license.
There's nothing in there that puts any restrictions on what you can do with the software all on your own, without sharing it with any friends. That doesn't mean that it's not under the GPL, just that the GPL doesn't apply.
Note, also, that if you were to download and modify a copy of a GPL program, modify it and then share it with one friend, you'd only be obligated by the license to share the source code with that one friend. If they then go and share it with a fourth party (1st party being original developer, you as second party, your friend is third party) you're not suddenly obligated to share the source code with that fourth party -- your friend is, however. (IOW, the distribution requirements don't work virally backwards up the chain)
I think Corel should, instead, have released it under the GPL with a strongly worded request that people not redistribute that Beta version and instead wait for the final version to come out. That way they wouldn't conflict with the GPL and they'd still manage to, largely, keep it from being distributed too widely.
Beta testing isn't "releasing" a product. I'd have to say that it does count as "distributing" a product, however, and that's what the GPL addresses, not "releases".
ProFTPD has been having a number of significant security problems. Otherwise it's a fine piece of software, though. I'd stay away from it for now.
h tml h tml h tml s ecurity-announce-17.txt
See:
http://freshmeat.net/news/1999/02/12/918825768.
http://freshmeat.net/news/1999/08/31/936140949.
http://freshmeat.net/news/1999/09/17/937609465.
http://www.suse.de/security/announcements/suse-
(note that SuSE has currently dropped it from their distribution.)
Personally, given that it's had more than one different security problem pop up in a relatively short time, I'd wait till they fix the existing bugs and then give it another month to see if any other bugs pop up before using it. (But, then, my need to worry at all about any kind of FTP service dissappeared last week and is unlikely to return.)
With credit card processing there is a bank network that you need to connect to somehow. The most traditional method is a modem built into a credit card terminal (what you see next to the cash register in brick-and-mortar stores).
Obviously, actually typing in credit card numbers manually is way too much work for an online store, but you still have to connect to that network somehow to do the processing.
Method #1: A modem hooked up to the computer. This is that CCVS and OpenCCVS like to do. (You could also do this with a leased line, but you'd need to be processing a *lot* of transactions to justify a leased line.)
Method #2: Connect over the internet to somebody else that does the connection to the "clearing house" for you. This is what CyberCash does. (CyberCash also has their own "cash" that customers can use, which is much better for small amounts of cash per purchase.)
I've dealt with CyberCash to some degree or another at two different jobs. Once you've got it set up, it's okay, as long as you don't have large volumes to work with. Cybercash has two basic problems. First off, it does everything over the internet, so it is at most as reliable as the internet. Secondly, it uses HTTP over TCP to perform the transactions. What this all means is that any kind of problems that occur will tend to leave you unsure whether or not the transaction has actually gone through or not. With older versions there was a database on your end that kept track of transactions that could get out of sync (with the fixes involving direct modification of the database) with their version of things. Newer versions of their protocol have "fixed" this problem by moving the database to their end. So if you deal with maybe one charge an hour (or 10 an hour, whatever), CyberCash will generally seem okay. When you have to deal with batches of about a thousand charges once a month, the problems start to become intolerable.
My suggestion: investigate the various options that involve your computer dialing the clearing house itself. Make *certain* that the software is capable of using a queue in an intelligent fashion (ie, if it has two charges waiting, dialing once to clear both), make sure the software has *some* way to handle more than one modem (if it comes to that) and try hard to find a bank that you can use ISDN with. (Since ISDN connects almost instantly)
(Note: I think Cybercash does have one competitor, but I don't think they support Linux yet and I suspect they'll have the same basic unreliability of the internet problems that Cybercash does)
- Create a customized RedHat installation.
- Mirror the RedHat distribution.
- copy the RPMs you want into the big RPMS directory, remove RPMS you don't want, etc. I'd highly reccomend replacing old RPMs with the updates and grabbing the ssh (etc) rpms from ftp.replay.com. Also, install autorpm and either customize it or make a separate RPM with just the files for autorpm to use an updates mirror that you make.
- Edit RedHat/base/comps -- the format is somewhat self-explanatory. For the purpose you're describing, go to the "0 --hide Workstation" section and customize that. It's probably easiest to make that list by doing a completely custom install off of the mirror you made in step one and then doing rpm -qa --queryformat '%{NAME}\n' to get a list of all the packages that got installed.
- run genhdlist on the appropriate directory.
- Put that all up someplace
- Optionally, do the whole kickstart thing.
- Set up an updates "mirror" someplace and point autorpm on all those workstations at that mirror. Presumably set up to only install from your relatively secure mirror and to check PGP signatures, etc...
You can put your custom install on an NFS mount point or burn it onto a CD (possibly even a bootable CD) and install all those machines... If you had more than 20 machines I'd suggest using DHCP, but you can probably manage without with 20 machines. You still might want to consider using DHCP to assign addresses to minimize how much effort you need to go through when you do the installs.Look on the GnuPG web page. There are links to a number of mail clients with some level of support.
Personally, I prefer mutt.