You mean “a variant of what FreeBSD has been doing since 2003”

Assuming we're only talking about ciphers and not protocols: by definition, there is one and only one possible ciphertext for any given combination of key and plaintext. Thus, there is no way to introduce a weakness in an implementation which would not be trivially detectable by comparison with any other implementation; in fact, the result would be unusable as it would not be interoperable with other implementations.

(With a caveat for algorithms which require a random initialization vector; don't let the implementation choose the IV for you.)

You can never be sure. The keystream of a good stream cipher is fully deterministic, yet statistically indistinguishable from the output of a good PRNG.

FreeBSD has been using Yarrow for 10+ years, and no FreeBSD release has ever shipped with the option to feed the stream from a HWRNG directly to /dev/random. The only news here is that we have a new framework in the kernel for plugging hardware pseudo-random number generators into Yarrow, and an explicit policy (issued in my capacity as FreeBSD Security Officer) to not expose HWPRNGs directly to userland. There was some pressure from corporate users who want the raw feed for compliance reasons, but they were told to use RDRAND etc. directly rather than through /dev/random.

s/a InfoCare/InfoCare/ obviously.

No, K is the SI unit for temperature, named after Lord Kelvin, who first suggested the concept of “absolute zero”.

You're the one who's confused here. The storage industry never “made their change”. They've always used powers of 10.

Assuming you are talking about MMU page sizes and not memory allocation: that may be true of the computer architectures with which you are familiar, but it is not universally true. The Sparc64 architecture, for instance, supports page sizes of 8 kiB, 64 kiB, 4 MiB, 256 MiB and 2 GiB. Older systems such as early Motoroal MMUs or early MIPS implementations had smaller page sizes (1 or 2 kiB).

Absolutely wrong. The use of kB to mean 1,024 bytes started around 1960, and only for memory. Bandwidth has always been, and is still, measured in powers of 10, not 2. Disk space was measured in powers of 10 until Microsoft came along and muddled the issue. Disk manufacturers still use powers of 10, like they always have. Software is a mixed bag, with some developers using powers of 10 and others using powers of 2.

Since someone mentioned Wikipedia: http://en.wikipedia.org/wiki/Timeline_of_binary_prefixes

In any case, a disk labeled 2 TB will never have exactly 2 TB or 2 TiB of storage space. The number on the label is just an approximation; the exact number is “as much as we can cram in and still have a reasonable amount left over for reallocation”.

HP usually resolve the issue (deliver parts and if necessary dispatch a tech) within the specified time frame. Dell rarely do. I haven't read the support contracts, so I don't know the details, but I _do_ know that the people who have negotiated those support contracts get royally pissed off when I tell them that Dell once again refused to replace a DIMM or disk on the basis of a predictive failure warning, or that our payroll database will have to run on only one server for a week while Dell scour warehouses on all five continents to scrounge up six DIMMs for the other.

I am in Norway as well. Dell subcontract the actual hands-on work to a InfoCare (as do HP, coincidentially), but handle all communication with the client themselves. They have a support center in Ireland staffed with techs from many different nationalities and generally try to route calls from Norway to Norwegian- or Swedish-speaking techs.

FWIW, we do most repairs ourselves, so the issue is “how fast can you send the parts” rather than “how fast can you dispatch a technician”. HP deliver most parts (disks, DIMMs, CPUs, RAID controller batteries) within a couple of hours but sometimes have to ship less common parts from other parts of the country or from Sweden. Dell deliver parts whenever they feel like it, which usually means within a day or two, but sometimes longer.

I work for an organization that has a large number of Dell servers, all of them with 5-year support contracts: a mix of 4-hour and next-business-day. In my experience, Dell have never, ever, ever solved an issue within the specified period of time. They also frequently refuse to replace failing parts until after they've actually failed (which AFAIK is a breach of the support contract), and they once told me that six DIMMs were a “large order” that would take a week to fill (after I'd already spent a week just getting them to agree that they needed replacing). They simply don't give a shit. I've had far better experiences with HP, but they also far more expensive.

IIUC this is one of the statements which the current lawsuits claims are materially false.

That may be open to interpretation, but "billed for work that was not performed" (IIUC he already sued her over those bills and won) and "stole jewelry from my house" is not. Those are statements of fact which she will have to either prove or retract.

Best summary I've read, by someone who is usually opposed to this kind of lawsuit but sees this particular case in a different light:

http://www.popehat.com/2012/12/06/yelping-about-bad-publicity/

It's the Norwegian equivalent. I wince every time I catch myself saying it.