Actually, under pressure from the SPA, the U.S. Congress has already banned renting software. However, despite the lobbying of Nintendo, video games are specifically excluded.
Before that, there actually were some retailers trying to make a go of renting software, but it didn't appear to be working too well. Scared the heck out of the software companies, though.
Since I used to use my Multias pretty heavily as web/mail/etc. servers, I collected a bunch of info about them here, including the service manual in PDF form.
Oh yeah, and if you don't think that $30 is a good deal for one of these, bear in mind that I paid an average of $1250 for mine.:-)
No, they aren't laptops, and they don't come with a monitor, keyboard, or anything else. It's a pizza box with an old, slow 166 MHz Alpha 21066 processor (think Pentium 100 class). But you have to add memory, disk drives, and whatever else you want.
The motherboard does have VGA video out, PS2 keyboard and mouse in, one parallel port, two serial ports, and two PCMCIA slots. There's an IDE connector for a 2.5 inch IDE drive. There's (barely) room for the 2.5 inch hard drive and a laptop (slim) floppy drive. Some models can accept a 3.5 inch hard drive, but then you lose the potential for PCI expansion. There is *no* room inside for an internal CD-ROM or any other drives. Also, the power supply isn't designed to power more disk drives. Don't even think about putting a 7200 RPM or 10000 RPM drive inside.
If you're lucky, the machine *might* come with a riser card that has a SCSI interface and a slot for a single PCI card.
If you're *really* lucky, the riser might even have an external SCSI cable.
Executive Summary: This is a good deal if you're familiar with Multias and don't mind scrounging the other needed pieces. But if you think it's a complete computer, forget it. Also, if you think it's some screaming fast Alpha, guess again.
"I want to pay for an operating system from a vendor with a contractual relationship that gives me recourse if things go wrong," Warnock said.
What OS does he think you can buy under a contractual relationship that gives him recourse? I quote from Microsoft's warranty for the OEM edition of Windows 98 as shipped with a Dell computer:
LIMITED WARRANTY. PC Manufacturer warrants that (a) the SOFTWARE will perform substantially in accordance with the accompanying written materials for a period of ninety (90) days from the date of receipt, [...]
So MS does in fact provide a warranty on the software, unlike many things for which you only get a warranty on the physical media. But the "accompanying written materials" consists only of a very slim booklet giving only the most general tutorial on the simplest features of the system. If anything more complicated is wrong, you're out of luck.
Suppose that you do have occasion to bring a warranty claim. Is Microsoft (or Dell) obliged to fix the problem? Heck no:
CUSTOMER REMEDIES. PC Manufacturer's and it's suppliers' entire liability and your exclusive remedy shall be, at the PC Manufacturer's option, either (a) return of the price paid, or (b) repair and replacement of the SOFTWARE or hardware that does not meet this Limited Warranty and which is returned to PC Manufacturer with a copy of your receipt.
I can't see where Warnock can really claim to have any substantive recourse with commercial software; AFAIK few if any OS vendors offer any better warranty than does Microsoft.
So in what way is the recourse with Linux any worse? If he acquires Linux over the net, at no charge, he has exactly the same option of getting a refund of the full price he paid for it, or a fix if the developers or community provide one. Exactly the same situation with MS, where MS gets to choose whether to refund the purchase price or provide a fix. MS's choice, not Warnock's. And even if he buys a commercially packaged Linux distribution, there are several that he can choose that offer a money-back guarantee.
The "contractual obligation of vendor" and "legal recourse" claims are widely cited as a problem with Free Software (or Open Source software, or...), but these claims are entirely without merit. It's a FUD tactic, plain and simple, and it's very disingeneous of Warnock to repeat such nonsense.
Anyway, if you've got the AirPort thing in your home, a cracker would likely have to be on your property if ot in your house to compromise it.
I live in a condo. I have at least twelve neighbors within a 150 ft. radius. Any one of them are in a position to try to break the thing without my knowing.
If someone other than my neighbors wanted to do it, there are plenty of places within that 150 ft. radius that someone could place equipment to monitor the relevant portion of the spectrum and forward the data elsewhere.
you can add extra encryption in the application layer.
That was my whole point. I was replying to someone who essentially claimed that 40-bit encryption is good enough. It isn't.
I think it would prolly be easier to physically break into your network than reliably and consistantly capture and decrypt AirPort transmissions.
Easier, definitely. But more likely to be detected. And with a higher risk of legal recourse.
Nowhere did I say that compromising the AirPort (or any other 802.11 system) would be easy. In the case of AirPort it would probably involve reverse engineering their firmware and using a hacked version to monitor the broadcast data stream (since presumably a standard AirPort won't provide direct access to the encrypted bits). If you think there aren't people out there capable of doing such a thing, you're very much mistaken.
It's not necessary for the attacker to personally have the skill to do this kind of stuff. All it takes is for one clever person to do it, and for the attacker to buy (or otherwise obtain) the resulting device.
Am I being paranoid? Yes, of course I am. How can you have any hope of maintaining a secure network unless you are paranoid about possible threat scenarios?
Well I don't think you have to worry *too* much about encryption for these puppies.
I beg to differ. I don't think you've spent much time thinking about threat scenarios.
any units not listening *specifically* for a message encrypted with that code will simply not "hear" it...
That may be enough to prevent someone from listening to an active TCP connection (assuming they can't hack a receiver), but there are definitely other threats. Two that come to mind:
I have a firewall between my ADSL and my home network. But with AirPort, there's only 40-bit encryption (weak) preventing an attacker from becoming part of my network. With my current wired Ethernet, the attacker would have to either physically break into my home, or subvert my firewall. It's always possible that the firewall has weaknesses that I'm unaware of, but unless the attacker knows or finds those weaknesses, he has to break a 168-bit triple DES key to attack it.
Even if an attacker doesn't want to break into my personal network, he can get free access to my ADSL internet connection. Again, assuming only that he breaks a 40-bit key.
I'm not saying that AirPort is a bad product. I'm only pointing out that you do have to worry about the encryption, and that 40 bits is pathetic. Obviously they wanted to avoid the stupid US crypto export regs. I'd be much more inclined to buy a unit with strong crypto, if some company starts selling them.
Not because I really think random people are trying to attack my network (although my firewall logs do show some script-kiddie attacks), but more as a matter of principle.
providing service for a given domain name need not incur any expense beyond a miniscule fraction of the annual operating costs of the registry and the root servers
Couldn't agree more. And in particular, the cost of maintaining a static entry in the database and root nameservers is essentially zero. I propose that a non-profit registrar should only charge for additions and changes to the information, with no per-year fee.
For those of you that aren't familiar with the details of DNS, this would not require you to pay a fee every time you add or rearrange hosts in your domain. It would only occur when you add or change name servers. In other words, only for the stuff that currently requires you to submit a template to NSI.
Of course, this won't work as long as the NSI owns the database and charges a per-zone annual fee to other registrars.
I think it is a crock that the NSI has been allowed to usurp ownership of the database. It should be owned by ICANN.
Note: I'm not necessarily a big fan of ICANN, but I think they're much less evil than NSI.
It's happening every day, all around us. NASA is unable to read much of their old data, and they've actually done more about preserving it than most organizations do.
It may seem obvious that a reasonable person would copy the data to new media every so often. But in reality, there's too much data to be copied, and the problem always seems like it can be pushed off to tomorrow. By the time tomorrow arrives, it's too late.
If you don't believe me, talk to any archivist. It's a difficult and thankless task.
why is keeping a hundred thousand copies of millions of pages journals on paper easier than on tape?
Magnetic tape stores data reliably for perhaps 10 years. Maybe 20 if it is kept under ideal conditions. From personal experience, some data recovery is still possible after 30 years, but not enough that you can count on it.
Paper lasts >100 years easily. If you go to the trouble of using acid-free paper, it lasts even longer.
On the one hand, it's important to have research results available in a convenient form today. But on the other hand, it still needs to be available many years into the future.
50 years from now [...] press a CD with a bunch of ASCII
Have you considered whether CD-ROM drives will be around 50 years from now?
It's already just about impossible to find anyone that can read data storage media that was common in the 1970s, such as 7-track tape.
When new media is introduced, it is common for new drives to read the previous generation media, such as DVD-ROM drives reading CD-ROM discs. But 50 years from now we will proabably have gone through at least three more generations of optical discs (or perhaps switched to an entirely different storage technology), so assuming that drives capable of reading CD-ROMs will still be readily available is very risky.
It is interesting to note, however, that when the CD data format was devised in the late 70s and early 80s, it was deliberately designed to be very simple to read, so that players would be inexpensive. The format is so simple, in fact, that players could be built without microprocessors. However, by the time the CD was introduced in 1983, microprocessers had become so inexpensive that AFAIK no microprocessorless players were ever marketed.
So even if CD-ROM drives aren't readily available 50 years from now, if someone is desperate enough to read one, it should be possible to build one. It seems like something a small team of graduate students could accomplish in less than a year (remember, they are assumed to have no existing CD hardware to work from).
That's what I would have thought. But looking at the data sheets does not reveal that the BR pins even connect to the chipset.
The FX chipset supported four-way SMP. I have not been able to find any technical reason why the LX and BX chipsets don't, other than Intel's assertion that they support two-way SMP.
Perhaps I'm dense, but I don't understand the concept of GPL'ing web pages. It's not software. What is the source code? What is the object code?
Is the HTML code the "source code"? If so, does that mean that if I print the page and give it to someone, they can demand that I give them the HTML code? So I'd be legally obligated to keep a copy of the HTML around (since it might no longer exist on the web site)?
I'm all in favor of GPL-like licenses, but in this case I think something else is needed.
This all makes sense with the "rules" that once a CSS encrypted data stream is decrypted, it can not be passed to anything but an output source. So, a hardware implementation of CSS couldn't pass unencrypted data back out the card to the computer or the card's data bus.
Nice theory, but it isn't true in practice. There are at least three ways to get the decrypted data out of the Ziva-DS on the Dxr2 card; one of them is even documented in the Ziva-DS manual: the DumpData_DVD command (0x0122). Another is to simply use the Ziva-DS' memory access feature to directly read the buffer that the decoded data is stored in; you just have to find the right address. The address isn't documented, and it changes with different microcode revisions, but it's pretty easy to find.
As for why they arent' releasing information on the decoder boards. It is mostly a legal issue. They have to protect their agreements to keep CSS implementation tightly gaurded.
Nice excuse, but it doesn't wash. They don't have to release any info on CSS to make it work, because the decryption is in hardware. OK, so you can't decode AC3 5.1 in software if you do that, but who cares? The built-in AC3 to Dolby Surround downmix in the Ziva-DS still works fine.
AFAIK, the Dxr2 software doesn't offer AC3 decoding anyhow.
No, they're being secretive because they (wrongly) believe that releasing the info would hurt their competitive position.
Yes, the existence of the NRO was declassified because they screwed up and put their new building in the unclassified budget. After that there's little point to pretending that the agency doesn't exist.
But it's still one of the most secret agencies whose existence is public knowledge.
I think you mean NRO (National Reconnaissance Office). If it weren't for the fact that their new building accidentally got listed on a public budget document (rather than buried in the black budget), most of Congress wouldn't even know that it exists.
I've had a Creative Encore 5x DVD package with the DXR-2 decoder board for a while and it works great. Why couldn't Creative just release Linux drivers for that board? It's already in production and it's not too expensive...
Sontas wrote:
Current generation hardware DVD decoding products and their related software do not have industry support in linux for at least one major reason. Most products out today perform the AC-3 decoding in software. This means that there is an associated CSS decryption happening in software.
Actually, most (if not all) of the hardware decoders products, like the aforementioned Creative Labs Dxr2, implement the CSS in hardware.
Early Dxr2 board had a Toshiba TC6807AF chip for CSS; newer boards omit that chip and perform the CSS function in microcode on the C-Cube ZiVA-DS chip (the actual MPEG decoder).
The main reason no one has released a Linux driver for the Dxr2 is not CSS, but because the available docs for the ZiVA-DS do not really contain sufficient detail on certain aspects of their operation, such as loading the microcode.
The vendors of MPEG-2/DVD decoder chips are being extremely secretive with their specs, for no good reason. An MPEG decoder is a very complex piece of hardware, but it has a relatively simple interface to the system. Just having the specs for the interface would not make it easy to clone the chip, but the marketing people don't seem to understand that.
I tried to convince LSI Logic to provide me with the necessary information to write a Linux driver for their L64021 decoder, and they were willing to discuss it, but they ultimately didn't do it. Sigh. (I wanted to use that chip because it can be interfaced to the VIP connector on many video cards, such as the Asus 3400 and 3800 Nvidia Riva TNT and TNT2 cards, and thus avoid the need for an analog VGA passthrough that degrades video quality.)
Kernels should remain in development until they reach true stability.
Sure, that's a nice idea. Motherhood, apple pie, etc. But you obviously aren't a real-world software developer.
What you propose won't work for several reasons:
Linus can't hold the development tree in a code freeze for the time it would take for the build to stabilize to the degree that you're asking for ("true stability"). If he tried, the various developers would fork off their own Linux kernels, and we'd have a big problem, worse than the egcs vs. gcc problem (which fortunately has been resolved).
If the kernel didn't get released to the "stable" branch at some point, it would never reach your desired level of "true stability", because not enough people would beat on it and find the bugs. Linus' policies are geared toward making sure that it seems pretty good before it is released to the stable branch, and then to shake out the remaining bugs.
Can you cite a single example of a software project of comparable complexity to the Linux 2.2 kernel which had fewer bugs at initial release? I didn't think so.
Actually, patching running operating systems used to be standard practice in the time-sharing days. Of course, you have to be very careful.
With Linux, just figure out where the offending instructions are by groveling through the compiler and linker output, and write to the relevant locations in/dev/kmem. For this particular bug, you probably only have to NOP out a few instructions.
Personally, I'm just as happy to reboot. It's not like it takes very long, and it's easier and safer. But if I were running a mission-critical 24x7 system, perhaps I'd think about it some more.
A segfault is something that happens to a process, usually due to a bug in user-space code. That process may have to be aborted, but the integrity of the kernel is not compromised.
A panic occurs when the kernel detects a condition that should never happen, and from which no good recovery is possible. It should not be possible to cause a panic from user-space code (except perhaps by root processes doing naughty things like scribbling on/dev/kmem).
Yes. But the point is that you don't want random people on the internet at large to be able to send bogus packets to your machine that cause it to panic.
Obviously there's no way to protect the machine against someone with superuser privileges from panicing it. But it is important to prevent unauthorized people from getting superuser privileges.
I'll be ordering a HP DAT24 DDS-3 drive next week so I can let you know how I get on
I've been using one for several months. It's great! I routinely write about 10G to the tape, and still have more than 1/3 of the tape to spare. The native capacity of DDS-3 is 12G, and they use compression to store more. Due to the nature of compression, your mileage may vary.
If you shop around on the net, you can find name-brand DDS-3 tapes for under $14.
Sony and HP have announced their DDS-4 drives. DDS-4 gets 20G native capacity. I was a little disappointed when DDS-4 was announced, as it is only a 67% capacity increase over DDS-3, while DDS-3 was a 200% increase over DDS-2. Anyhow, as near as I can tell, Sony is shipping their DDS-4 drive already but HP is not. I haven't seen any DDS-4 tapes at retail yet.
Some people whine about the cost of DDS drives. It's almost irrelevant to me; I spend much more money on tapes. I find it to be a much better tradeoff to buy an expensive drive that uses inexpensive tapes. Most of the Travan and similar things are priced the other way around, so I have no use for them.
Unfortunately only the HP-71B was IEEE-854 compliant. Calculators such as the 28, 38, and 48 are not. However, the arithmetic routines were derived from those of the 71, so they have some of the characteristics of IEEE-854. They've omitted the NaNs, infinities, rounding modes, denormalized numbers, and such.
Slashdot Mirror
Before that, there actually were some retailers trying to make a go of renting software, but it didn't appear to be working too well. Scared the heck out of the software companies, though.
The Multia also has built-in 10 Mbps Ethernet, with 10-base-T (twisted pair), 10-base-2 (thin coax), and AUI connectors.
Some Multia owners have experience reliability problems which can be solved by replacing a chip and/or the fan. Details are in the Net/BSD alpha Multia Frequently Asked Questions.
Since I used to use my Multias pretty heavily as web/mail/etc. servers, I collected a bunch of info about them here, including the service manual in PDF form.
Oh yeah, and if you don't think that $30 is a good deal for one of these, bear in mind that I paid an average of $1250 for mine. :-)
The motherboard does have VGA video out, PS2 keyboard and mouse in, one parallel port, two serial ports, and two PCMCIA slots. There's an IDE connector for a 2.5 inch IDE drive. There's (barely) room for the 2.5 inch hard drive and a laptop (slim) floppy drive. Some models can accept a 3.5 inch hard drive, but then you lose the potential for PCI expansion. There is *no* room inside for an internal CD-ROM or any other drives. Also, the power supply isn't designed to power more disk drives. Don't even think about putting a 7200 RPM or 10000 RPM drive inside.
If you're lucky, the machine *might* come with a riser card that has a SCSI interface and a slot for a single PCI card.
If you're *really* lucky, the riser might even have an external SCSI cable.
Executive Summary:
This is a good deal if you're familiar with Multias and don't mind scrounging the other needed pieces. But if you think it's a complete computer, forget it. Also, if you think it's some screaming fast Alpha, guess again.
Eric (owner of two 233 MHz VX42 Multias)
Suppose that you do have occasion to bring a warranty claim. Is Microsoft (or Dell) obliged to fix the problem? Heck no:
I can't see where Warnock can really claim to have any substantive recourse with commercial software; AFAIK few if any OS vendors offer any better warranty than does Microsoft.So in what way is the recourse with Linux any worse? If he acquires Linux over the net, at no charge, he has exactly the same option of getting a refund of the full price he paid for it, or a fix if the developers or community provide one. Exactly the same situation with MS, where MS gets to choose whether to refund the purchase price or provide a fix. MS's choice, not Warnock's. And even if he buys a commercially packaged Linux distribution, there are several that he can choose that offer a money-back guarantee.
The "contractual obligation of vendor" and "legal recourse" claims are widely cited as a problem with Free Software (or Open Source software, or ...), but these claims are entirely without merit. It's a FUD tactic, plain and simple, and it's very disingeneous of Warnock to repeat such nonsense.
If someone other than my neighbors wanted to do it, there are plenty of places within that 150 ft. radius that someone could place equipment to monitor the relevant portion of the spectrum and forward the data elsewhere.
That was my whole point. I was replying to someone who essentially claimed that 40-bit encryption is good enough. It isn't. Easier, definitely. But more likely to be detected. And with a higher risk of legal recourse.Nowhere did I say that compromising the AirPort (or any other 802.11 system) would be easy. In the case of AirPort it would probably involve reverse engineering their firmware and using a hacked version to monitor the broadcast data stream (since presumably a standard AirPort won't provide direct access to the encrypted bits). If you think there aren't people out there capable of doing such a thing, you're very much mistaken.
It's not necessary for the attacker to personally have the skill to do this kind of stuff. All it takes is for one clever person to do it, and for the attacker to buy (or otherwise obtain) the resulting device.
Am I being paranoid? Yes, of course I am. How can you have any hope of maintaining a secure network unless you are paranoid about possible threat scenarios?
- I have a firewall between my ADSL and my home network. But with AirPort, there's only 40-bit encryption (weak) preventing an attacker from becoming part of my network. With my current wired Ethernet, the attacker would have to either physically break into my home, or subvert my firewall. It's always possible that the firewall has weaknesses that I'm unaware of, but unless the attacker knows or finds those weaknesses, he has to break a 168-bit triple DES key to attack it.
- Even if an attacker doesn't want to break into my personal network, he can get free access to my ADSL internet connection. Again, assuming only that he breaks a 40-bit key.
I'm not saying that AirPort is a bad product. I'm only pointing out that you do have to worry about the encryption, and that 40 bits is pathetic. Obviously they wanted to avoid the stupid US crypto export regs. I'd be much more inclined to buy a unit with strong crypto, if some company starts selling them.Not because I really think random people are trying to attack my network (although my firewall logs do show some script-kiddie attacks), but more as a matter of principle.
I never understood why there was a big flap over that quote. Math is hard. I failed diffeq.
For those of you that aren't familiar with the details of DNS, this would not require you to pay a fee every time you add or rearrange hosts in your domain. It would only occur when you add or change name servers. In other words, only for the stuff that currently requires you to submit a template to NSI.
Of course, this won't work as long as the NSI owns the database and charges a per-zone annual fee to other registrars.
I think it is a crock that the NSI has been allowed to usurp ownership of the database. It should be owned by ICANN.
Note: I'm not necessarily a big fan of ICANN, but I think they're much less evil than NSI.
It may seem obvious that a reasonable person would copy the data to new media every so often. But in reality, there's too much data to be copied, and the problem always seems like it can be pushed off to tomorrow. By the time tomorrow arrives, it's too late.
If you don't believe me, talk to any archivist. It's a difficult and thankless task.
Paper lasts >100 years easily. If you go to the trouble of using acid-free paper, it lasts even longer.
On the one hand, it's important to have research results available in a convenient form today. But on the other hand, it still needs to be available many years into the future.
It's already just about impossible to find anyone that can read data storage media that was common in the 1970s, such as 7-track tape.
When new media is introduced, it is common for new drives to read the previous generation media, such as DVD-ROM drives reading CD-ROM discs. But 50 years from now we will proabably have gone through at least three more generations of optical discs (or perhaps switched to an entirely different storage technology), so assuming that drives capable of reading CD-ROMs will still be readily available is very risky.
It is interesting to note, however, that when the CD data format was devised in the late 70s and early 80s, it was deliberately designed to be very simple to read, so that players would be inexpensive. The format is so simple, in fact, that players could be built without microprocessors. However, by the time the CD was introduced in 1983, microprocessers had become so inexpensive that AFAIK no microprocessorless players were ever marketed.
So even if CD-ROM drives aren't readily available 50 years from now, if someone is desperate enough to read one, it should be possible to build one. It seems like something a small team of graduate students could accomplish in less than a year (remember, they are assumed to have no existing CD hardware to work from).
The FX chipset supported four-way SMP. I have not been able to find any technical reason why the LX and BX chipsets don't, other than Intel's assertion that they support two-way SMP.
Is the HTML code the "source code"? If so, does that mean that if I print the page and give it to someone, they can demand that I give them the HTML code? So I'd be legally obligated to keep a copy of the HTML around (since it might no longer exist on the web site)?
I'm all in favor of GPL-like licenses, but in this case I think something else is needed.
AFAIK, the Dxr2 software doesn't offer AC3 decoding anyhow.
No, they're being secretive because they (wrongly) believe that releasing the info would hurt their competitive position.
But it's still one of the most secret agencies whose existence is public knowledge.
Where's the NRO home page? :-)
Sontas wrote:
Actually, most (if not all) of the hardware decoders products, like the aforementioned Creative Labs Dxr2, implement the CSS in hardware.Early Dxr2 board had a Toshiba TC6807AF chip for CSS; newer boards omit that chip and perform the CSS function in microcode on the C-Cube ZiVA-DS chip (the actual MPEG decoder).
The main reason no one has released a Linux driver for the Dxr2 is not CSS, but because the available docs for the ZiVA-DS do not really contain sufficient detail on certain aspects of their operation, such as loading the microcode.
The vendors of MPEG-2/DVD decoder chips are being extremely secretive with their specs, for no good reason. An MPEG decoder is a very complex piece of hardware, but it has a relatively simple interface to the system. Just having the specs for the interface would not make it easy to clone the chip, but the marketing people don't seem to understand that.
I tried to convince LSI Logic to provide me with the necessary information to write a Linux driver for their L64021 decoder, and they were willing to discuss it, but they ultimately didn't do it. Sigh. (I wanted to use that chip because it can be interfaced to the VIP connector on many video cards, such as the Asus 3400 and 3800 Nvidia Riva TNT and TNT2 cards, and thus avoid the need for an analog VGA passthrough that degrades video quality.)
What you propose won't work for several reasons:
Can you cite a single example of a software project of comparable complexity to the Linux 2.2 kernel which had fewer bugs at initial release? I didn't think so.
With Linux, just figure out where the offending instructions are by groveling through the compiler and linker output, and write to the relevant locations in /dev/kmem. For this particular bug, you probably only have to NOP out a few instructions.
Personally, I'm just as happy to reboot. It's not like it takes very long, and it's easier and safer. But if I were running a mission-critical 24x7 system, perhaps I'd think about it some more.
A panic occurs when the kernel detects a condition that should never happen, and from which no good recovery is possible. It should not be possible to cause a panic from user-space code (except perhaps by root processes doing naughty things like scribbling on /dev/kmem).
Obviously there's no way to protect the machine against someone with superuser privileges from panicing it. But it is important to prevent unauthorized people from getting superuser privileges.
If you shop around on the net, you can find name-brand DDS-3 tapes for under $14.
Sony and HP have announced their DDS-4 drives. DDS-4 gets 20G native capacity. I was a little disappointed when DDS-4 was announced, as it is only a 67% capacity increase over DDS-3, while DDS-3 was a 200% increase over DDS-2. Anyhow, as near as I can tell, Sony is shipping their DDS-4 drive already but HP is not. I haven't seen any DDS-4 tapes at retail yet.
Some people whine about the cost of DDS drives. It's almost irrelevant to me; I spend much more money on tapes. I find it to be a much better tradeoff to buy an expensive drive that uses inexpensive tapes. Most of the Travan and similar things are priced the other way around, so I have no use for them.
Unfortunately only the HP-71B was IEEE-854 compliant. Calculators such as the 28, 38, and 48 are not. However, the arithmetic routines were derived from those of the 71, so they have some of the characteristics of IEEE-854. They've omitted the NaNs, infinities, rounding modes, denormalized numbers, and such.