It's now been nearly two years since I was diagnosed with coeliac disease. In that time I've mostly come to terms with the required changes in my diet. I've adapted to a life without pizza; I've either learned to cook, or found gluten-free varieties of, the Chinese meals I missed most; I've done the minimal necessary adaptation of my usual bulk self-catering habits; and generally, apart from it being nearly impossible to eat out and often severely inconvenient to have friends cook for me, I basically don't feel too annoyed most of the time by the whole business.

One thing I have had trouble finding, though, is good gluten-free bread. I have several options which are suitable for making into breadcrumbs for stuffing and bread sauce, and one or two which make passable toast if you lay on pâté with a trowel, but last week I would have told you I knew of no GF bread products which I'd be willing to eat in the form of actual bread.

Within the last few days I've been fed three startlingly good GF bread products, all of which were made by Dietary Specials, a company which I previously knew for making decent GF steak pies and utterly inedible GF ready-meal pizzas. I don't know if any of them would be good enough to adopt as a regular part of my diet, but for the occasional sausage butty or light snack they're definitely an improvement on anything I'd previously encountered. I will have to go and buy some more.

The other day I found a cheap copy of Arthur C Clarke's ‘2010: Odyssey Two’ in a charity shop. My bookshelf was lacking a copy and I hadn't read it in years, so I snapped it up and re-read it.

At one point it describes the means by which the people on the spaceship secure their communications links back home:

[…] the cipher was based on the product of two hundred-digit prime numbers, and the National Security Agency had staked its reputation on the claim that the fastest computer in existence could not crack it before the Big Crunch at the end of the Universe.

It doesn't take more than a nodding acquaintance with cryptography to recognise the cipher being described here as RSA; but two prime numbers of 100 decimal digits each come to what in modern terminology we would usually describe as a 665-bit RSA key, and that's frankly feeble by modern standards (768-bit keys are a thing of the past and 1024-bit keys are starting to look shaky). For once, Clarke's vaunted foresight seemed to have let him down.

Then I stopped and thought a bit harder. He was writing this around 1980, trying to look thirty years into the future, and not only has he got the encryption algorithm right (well, probably – it isn't 2010 yet and there's still time for surprises, but it doesn't currently look as if RSA will be generally abandoned by then), but he's also within a factor of ten of the right key length – and over a thirty-year period, to be out by even a full factor of ten only needs a 10% error in the doubling period of Moore's Law (although I'd guess it's more likely that what Clarke really failed to anticipate was improved factoring algorithms). So, actually, that's not bad at all!

In the last couple of days I've been updating the PuTTY online help mechanism in anticipation of Windows Vista discontinuing the old-style Windows Help. I've prepared a new-style HTML help file, but I want to keep the option to use the old one, because (among other reasons) PuTTY still tries to work on Win95, which doesn't support new HTML help. So I hacked together an automatic detection mechanism which will use the HTML help file if it's both present and supported by the OS, and will fall back to WinHelp otherwise.

In order to make sure this really worked on Win95, I needed a Win95 machine to test it on. The PuTTY team used to have a real Win95 machine available, but currently doesn't; so lacking any other convenient options, I dug out my old Win95 install media from a dusty pile of CDs (amazingly, even the boot floppy which accompanied the CD was in full working order) and attempted to install it on a Bochs virtual machine.

It took me twenty minutes of editing configuration files and installing extra Debian packages to even start Bochs; no individual hoop I had to jump through seemed actually unreasonable, but it all added up to a lot of hassle. Eventually I had a working virtual machine and was able to run the install process – which ran very smoothly and promisingly up until the install program said ‘Setup is preparing to install files’ and then sat there beating its little drum icon for over an hour until I lost patience and killed it.

So then I tried QEMU instead, and was pleasantly surprised to find that it didn't even need a configuration file. I gave it command-line arguments pointing it at a floppy image, a hard disk image, and a CD image, and let it rip; it just worked, it ran much faster than Bochs, it installed without a hitch, and it's now cheerfully running a working Win95 system for me on which I've been able to test the help file switching mechanism and confirm that it works fine.

Somebody told me once that if you buy a medium-price hi-fi amplifier, it will come with tone controls, graphic equalisers, and no end of knobs and dials on the front so you can adjust it to exactly the sound you want, but if you buy a really expensive one it will just have a power switch and a volume control and get everything else right without having to be told. That was the feeling I got from comparing Bochs to QEMU.

(Unfortunately, it all went a bit pear-shaped when I tried to get the networking to work; I managed it in the end, but it wasn't nearly as painless as the rest of the process. I think this was mostly Win95's fault, however.)