On the other hand, they are almost perversely non-random-access(doubly so if you are talking about a multi-tape set with library swapping, more than doubly so if you are talking about a multi-tape set larger than your library can handle with junior-admin swapping), so the 'somebody fucked up, they need File X to be like it was three months ago' scenario sucks.
Nearline or consumer SATA on undistinguished controllers aren't quite as zippy; but they might as well be a RAMdisk compared to tape if you are doing small-scale restores, though probably not as fast if de-icing a much larger dataset for wholesale restoration of multiple systems.
Now, if you just need a few megs of cheap storage and don't want to bump the BoM, building an arbitrary filesystem access mechanism seems so sloppy and unconcerned with actual security as to make me wonder what else they fucked up; but SELinux is pretty powerful, if a pain, at granular lockdown of lousy or dangerous software.
Achieving a name recognized by somebody other than you is a somewhat more complex problem, usually requiring a certain amount of give-and-take in terms of "I'll accept your stupid idea if you endure mine" type arrangements.
For all the histrionics about it, Nobody was somehow magically anointed the Super Name Czar by some magically authoritative process. Some organizations have their shit together, and any names in a given domain not endorsed by them are pretty much just private nicknames, some don't; but that's it.
An actually contemporary tape drive(and a machine capable of keeping it fed when it is running full bore) is Not Cheap; but the fleabay shit that is cheap tends to offer painfully mediocre capacity and unknown reliability. Disks, by contrast, have a cost of entry that basically starts at zero and scales more or less linearly with the number of disks, unless you absolutely must have them all online at the same time(and even here, you hardly need screamin' hardware RAID for your backup volume, and bulk SATA ports of undistinguished performance are cheap).
It's a pity that they apparently didn't have the tools to distinguish the wreckage of a white aircraft from snowcap at that time (or in that place); but their odds of having something worth finding were a lot better.
That, to the best of my understanding, is what confuses people: Qt is generally considered superior to Gtk, is now LGPL, and is generally well liked; so why would somebody say 'Well, Gtk has issues, so I'm going to make my own Gtk; but better!' when they could just use Qt?
It's been what, over three days now, with an aircraft that disappeared from radar at commercial cruising altitude without so much as a burst of garbled obscenities from the flight crew. Do you think that your family is clinging to those little flotation-device pillows, awaiting a rescue that would have come in time if only for upgraded real-time blackbox transmission?
If anybody derives some sort of comfort from whatever they do manage to find, all the better; but this is all trying to recover data for failure analysis, not survivors.
Now, if you want to justify real-time transmission, check out the amount of (incidentally not paid for by the airline) search gear that has been diverted from Malaysian, Chinese, and other sources to looking for the debris. Whole bunch of ships, airplane and helicopter overflights, diversion of what, 10 satellites? That starts to make the $100k look like savings.
Yet if you point this fact out, you lose a presidential election...
Actually, there may be a confounding factor (this is a hypothesis, anybody who has real numbers is welcome to step forward to argue for or against):
Assuming you aren't inimically opposed to the concept of the welfare state (whether because you think that it's actually a good idea, or whether you think that it's relatively cheap insurance to keep a fundamentally capitalist economy with slightly higher tax rates to buy off the proles, irrelevant), the state basically has two options for 'redistribution':
1. Actually comparatively high taxes on individuals and corporations, used to fund a variety of not-directly-cash public services(eg. national health system, cheap or free education, etc.).
2. Avoid the flack associated (in the US) with robust public-sector offerings, and sneak in your social welfare spending primarily in 'emergency' programs (WIC, etc. which pay in scrip; but have nontrivial USD value once you discount them for being able to purchase only certain classes of goods) and in 'hand up for the virtuous poor' type things ("earned income tax credit", assorted subsidies for small business loans, edging up to programs that are basically a sop for the middle class, like mortage related deductions).
Now, lest anybody misinterpret me on this point: I Think It Is A Bad, Bad, thing that nontrivial swaths of the US population are basically so damn poor that the only cash worth squeezing out of them is sales taxes and check-cashing joint fees. However, barring a solution to that problem, it would be my contention that (like our absurd 'We should really have universal health care; because our current system is an utter clusterfuck delivering bad results for crazy high prices, and tying workers to their jobs; but universal health care is commie socialism, so let's have a crazy arrangement where the government 'launders' universal health care(at a tidy markup) through the incumbent private insurance companies!') our 'let's see if we can get some of the benefits of a welfare state without courting the unpopularity of calling it that, and without the clout to do anything about the ever-widening wealth gap' approach has left us with a singularly dysfunctional creature, neither fish nor fowl.
Some of them also do independent work; but others (in terms of customer base and income) are basically federal employees in all but name and price.