It would be great if it were to accidentally succeed in its stated mission -- no one would complain -- but it's about H1-Bs now, not the future of America's children.
I realize that some H1-B workers are able subsequently to immigrate, but it's a separate route and it's not the program's intention. It's good to have a program that lets highly specialized advisors in -- it ensures knowledge and skills transfer from the broader world, but H1-B is not primarily used for that. The program would have more value if it had higher standards for quality, not larger quantity. As it stands, it's simply an attack on labour, one that cloaks itself in the language of freedom and immigration while providing neither.
Couldn't one keep the development boxes remoted somewhere, and just access them through remote terminal over LTE? The latency is okay, and even a lot of remote access isn't going to blow through a 15GB data plan.
CF cards are IDE, but with a smaller pin-out. If you have an adapter between the laptop IDE and the CF form factor you might be able to either plug the HD into a newer box with a CF adapter or plug a CF card directly into the laptop (assuming there's a second slot... or possibly even slaved on to a single cable if there isn't).
Personally I'd try PCMCIA ethernet because I still have a card or two in my basement, but who knows what crap you have.
Really, though, I just want to say thank you to the poster for a problem that Slashdotters really care about.
If any of you kiddies are interested in technology the NSA will have trouble getting at, I know of someone with a Contura laptop to sell you...
Shannon's theorem is true for a single channel. Eventually, cramming in more bits into one channel becomes power-prohibitive, because one must double power for each new bit added in. The benefits from adding power diminish even further when a system is widely deployed, as power from one system shows up as noise in the next, and SNR in all systems hits an interference limit.
To get around these limits, two related techniques are used
1) adding more antennas, to create more channels which are separated in space
2) coordination techniques that reduce interference from one spatial channel to another
If 2) is done well, then 1) can provide the kind of linear benefits you'd hope for - each new channel contributes its share to the sum data-rate. As you might imagine, building very parallel radio systems and coordinating what's sent over them presents its share of challenges.
For those of you looking to try it at home, there is at least one Software Defined Radio platform that will drive the ~1000 spatial streams you would need for this:
"leaving law enforcement a way in" like a rubber hose?
My choice of "timebomb" was poor. I meant only that something complex, valuable, and easy to connect to would be in danger of getting compromised, and that being able to receive patches OTA would mitigate this threat better if it didn't make the thing even easier to connect to.
There is some risk of seeing manufacturers ship (literally) cars that are half-baked, but there are still consequences to messing up. While the prohibitive costs of a recall force some more attention to detail during design, they also can act to discourage manufacturers from acknowledging and fixing things. There's moral hazard either way -- it's difficult to design one's way out of sloth and risk. From a security perspective, cost / benefit analysis and "appropriate" security is often emphasized over defense in depth, so there's risk that resources spent on, eg, private cellular access are resources taken away from other system hardening efforts rather than something layered on top. It's often the case that the defender isn't really playing to win.