dichro: (Default)
I was pretty excited when 23andMe first appeared on the scene, offering genomic scanning to the masses. In the years since, beyond distributing several 23andMe kits as gifts and introducing people to Promethease, nothing much happened. The whole FDA debacle didn't help. But after we decided that we wanted to have children, my interest surged again, and right about the same time, Genos Research began (briefly) to offer "whole exome scanning", where the exome is the part of the genome that codes for proteins, and constitutes about 50M basepairs, compared to most other services' microarray-based scanning that covers <1M.

Naturally, I signed us up.

Read more... ) At any rate, after much banging and cursing, I do have something that generates superficially interesting data for potential progeny, given a pair of genome scans from Genos or 23andMe. Try it out?
dichro: (Default)
After agonizing about it for a while, I replaced Grace's door locks with Kwikset Kevo 2s. This wasn't just technology for technology's sake: I wear a backpack to work and generally lack pockets, and thus digging out a key is a hassle both leaving and returning. But convenience ever trades off security, and I felt conflicted about the idea for a long time.

The conflict wasn't, I think, really about security risks; it was that, in making the decision to change out the locks, I become accountable for all consequences. I've been burgled before, and even though I wasn't the one who was present at the time, it dramatically changed how I felt about my home. If that happens again, I'm always going to wonder if the old locks would have prevented it, particularly given how emotionally invested we are in Grace.

But realistically, the question of objective impact on security is quite separate from the psychological aversion to culpability, and my best assessment of the former is that it doesn't make much difference. While installing smart locks indubitably increases the attack surface, the external locks themselves are far from Grace's point of greatest vulnerability. We have huge and easily accessible windows. The door-frames appear to be neither reinforced nor particularly rigid. There's no Bluetooth repeater installed that might allow remote access (at the same time, we can't remotely verify that we locked our doors, which I do get very occasional OCD tingles about). Our security system shares no protocols (and therefore no vulnerabilities) with the locks. So while someone could bring along an RF amplifier or Bluetooth hacking rig to break in, that seems like a lot of unnecessary complexity when a crowbar would do the job just as fast.

At any rate, having decided to do it, installation wasn't complicated. The only surprises were, firstly, that lolcontractor had butchered the doors when installing the original locks and then tried to compensate with various forms of reinforcement that needed working around, and secondly that our front door is 2⅛" thick, and Kwikset claim not to make a thick door kit for the Kevo. This is a half-truth: the Kwikset 84221 thick door kit's bolts fit the Kevo, and even though the kit's tailpiece doesn't, the Kevo's own tailpiece slides out far enough to work unmodified.

At the end of it all, things are working well. The original keys still work (since the original locks also happened to be from Kwikset's self-rekeyable line), but I don't use them unless my phone is out of battery. Kevo's Bluetooth security appears robust, which is more than you can say for its physical security, but then see above re: physical bypass not being the thing this lock needs to protect against. So far, I'm pretty happy with the situation.
dichro: (Default)
Grace is a mess of incomplete projects. I mostly blame the contractor who did the renovation before us for laziness (ha), though I guess I should also be better about seeing all of these as opportunities.

The recent success is finally fixing the Wi-Fi coverage in the library. The contractor didn't deign to put in conduit in the walls... or any kind of reasonable cabling, either. There is, at least, a wallplate in every room with a phone jack and coax, but all of the cabling is cat3, stapled to various beams along the way. Most of it is only connected at one end, too, with the other end dangling loosely in the attic or wall cavity, so getting any kind of aesthetically appealing wired connection between rooms has been challenging.

The DSL link is terminated downstairs, at a wall socket with two other cat3 cables present. I'd earlier worked out that one of those went up to the attic, and I repurposed a ceiling mounted smoke alarm's remodeling box to connect a Unifi AP. That mostly covers the top floor, and another AP directly connected near the DSL model handles the basement, but the signal at the main floor at the other end of the house is too weak to be able to video-conference from the couches in the library, which is something of a pity, given how beautiful that room is and my love for showing it off.

There's an enclosed crawlspace below the library which I've long had in mind as a machine room (despite an unhealthy amount of dust left over from construction), but I hadn't found a cabled path to it, despite one clearly visible cat3 cable. Two cable tracers and one unattractive patch cable later, it's finally wired up, with a third Unifi perched on a beam below the library (there is a wall plate in the library, presumably with a cat3 connection, but I have no idea where it connects to, and we bolted bookcases onto that wall anyway).

Together with a Unifi PoE switch and a Unifi Security Gateway, Grace is now an all-Unifi network which I manage by clicking around a UI on a controller that I don't even host and I just don't recognize myself any more. Somewhat to my surprise, all of the links are running cleanly at 1GbE. Modern ethernet hardware is pretty amazing.

After some prompting, we bought into the whole Sonos thing for sound. Even the Play:1s have great audio quality, but after investing in a bunch of them and a Playbar, we discovered that they don't work as simple wireless speakers. In particular, Eden wants to play audiobooks and movie soundtracks through them, and... can't. They just don't work that way. Which is amazing, but there you go. As a last ditch effort before investing in yet another piece of Sonos kit that was sufficiently expensive to offer external input to a standalone Bluetooth audio receiver (I can't even), I tried airsonos, which fakes up a bridge between AirPlay and Sonos hardware, and it sort of works. It only really solves half the problem, though, since it adds close to 10 seconds of audio latency, which means audiobooks are fine, but movies are not. I have some small hopes of this being solved for me, but since I also lack a host to run airsonos permanently, it's moot for the moment.

There's yet more ecosystems to kvetch about, but I'm realizing that about 90% of my complaints are about the wiring in the house. I was always politely bemused by all the earnest threads about choice of conduit when renovating or building. Now I finally understand. Argh.
dichro: (Default)
 After I threw up a gallery with empty pedestals available for anyone to place art on, I felt a little frustrated, maybe more than a little bored. Clearly, nobody was actually going to do that, because why? After a couple of weeks, I decided to crawl the internet and collect art instead. Now that that is done, and intangible.gallery is live, I'm back at the frustrated and bored stage, though at least with more to show for it. I could:
  • Add SSL everywhere, including client-certificates for authentication, as a grounding for RPC authorization, and maybe do something clever with having people rotating clearly mis-posed objects in the gallery, or doing a robots.txt-style ownership claim to their own art.
  • Attack the rendering glitches that come with simultaneously runtime-processing complicated meshes and textures. Maybe hang out with the unity-savvy folk on forums some more too.
  • Finish the last little pieces of the current iteration, surfacing object stats and licensing information to viewers, broadening the scope of the crawler, handling more object and texture formats, all the endless tweaks.
But mostly, I'm just a little bored. I had a brief frisson of excitement when friends tried it out, so maybe advertising the gallery's existence more widely is the best path to regaining my enthusiasm. In the meantime, I could also wrap up messing around with smart door locks...

dichro: (Default)
For a project that's a thin layer of glue [sic] on top of user-supplied RPCs, calling RPCs sure is hard. I mean, discovering RPCs is even harder, but unfazed, I added empty art pedestals to the Intangible Gallery, with APIs to allow all and sundry to place objects on them.
Up until a few minutes ago, I thought that the SteamVR overlay keyboard didn't offer symbols, which was going to make typing in URLs for meshes and textures challengingThen I accidentally squeezed the grip buttons on the controller, and voila: symbols appeared! So, yeah, discoverability.

Anyway. I was going to add some icons to the controllers to help with that discoverability thing, but saved it for next version. For the moment, one-way RPC calls are sort-of complete-ish, in the sense that a single-layer record for an RPC request can be populated by the latest version of the client (painfully, using either on-screen or physical keyboards), dispatched, and the text of any reply received is then crudely displayed.

These RPCs are defined using the Avro schema language, which is faintly troubling, since I'm resolutely sticking to protobuf for the streaming presence updates between client and server. I rationalize it like so: the client-server protocol is purely for the machines, supporting what's basically a distributed cache-coherence exercise. Messing with that protocol is unlikely to improve user experience, and it's the highest-volume thing going on, so it's optimized for speed and predictability at the expense of human-readability, ergo protobuf. The client-client protocol, OTOH, is all about personalized tweaks, messing around, version skews, etc; and much less performance-sensitive. Hence, Avro.

Speaking of the client-server protocol, it needs some (compatibility breaking) changes too. Right now, everything assumes that updates are complete, rather than incremental. There's a few useful changes to be made here:
  1. Separate client update type from server update type. Limit the kinds of changes that clients can make via the streaming protocol (ie, only send pose updates).
  2. Provide a registration RPC for clients to call, supplying avatar representation, RPCs offered, etc. (For bonus points, consider whether the server could offer an RPC proxy for the client, and what that could offer (client privacy; positive affirmation of caller presence and VR context))
  3. Nail down semantics of incremental updates.
  4. Be more careful with default values. eg, prefer zero-based rescale (-1, ∞) rather than one-based scale (0, ∞), so that the default does something useful.
dichro: (Default)
One score and negative one years ago, I read a book titled Snow Crash, by one Neal Stephenson. Notably, it featured the Street, aka the Metaverse, a virtual environment where diverse characters meet and do unto one another. I was cutting my teeth on LPMuds at the time, and unreasonably fascinated by the Java VM, which led to a terrible idea: I was going to write the Metaverse.

Writing an environment to handle mutually hostile foreign code was briefly fun but left me rather jaded, and I abandoned the project; neither the first nor the last to tilt at that windmill. I have come back to it several times since, generally with the same (lack of) result. 

My most recent attempt started earlier this year, when Amazon released Lumberyard, a 3D game engine. On poking around, I was puzzled to discover licensing restrictions that forbade use of any cloud services with Lumberyard other than AWS. The epiphany came slightly later: "the cloud" really might change everything.

(Disclaimer: I was kinda late to all the cloud hype. I'm still not totally convinced by this whole new-fangled web thing.)

All of my previous ideas for writing the Metaverse revolved around uploading code to some central service that would run it for you. I thought I was being clever by having users upload JVM bytecode, rather than giving them another language to learn. But Cloud services are falling all over themselves to make it as easy as possible for users to write and host their own publicly accessible code. A shared virtual environment can be a thin layer of glue on top of that. That on its own was a pretty exciting idea, but when I tried out a friend's Vive, I was hooked.

I published the first version of an example server (source) and client (binary) a few days ago. It happened to be exactly 25 years since a chap named Tim did the same thing with his little project.
dichro: (Default)
One of my pre-roadtrip notions was to broadcast an APRS beacon from the car, for tracking our progress across the country. Having a radio in the car, I further justified, was also a useful emergency communications tool for, you know, emergencies. Without further ado, I bought a Yaesu VX-8DR, advertised as an APRS/GPS/Bluetooth radio.

This description turned out to be taking certain liberties with the truth.

It did Bluetooth in the sense that you could buy a separate $80 Bluetooth radio to plug into it.
It did GPS in the sense that you could buy a separate $75 GPS radio to plug into it (with an additional $40 adapter).
It did APRS, to be sure, in that it had a built-in 1200/9600 baud modem, which must be, what, a hundred lines of code to replicate in software?

Affronted, I refused to spend the money, deciding that I had a perfectly good GPS in my phone that is running ~100% of the time in the car anyway, and there's no reason to buy all that extra hardware when I could just send it to the radio's serial port via a $11 Bluetooth adapter. There's even a range of GPS mouse apps out there ("GPS mouse" being a term that made a lot more sense before Bluetooth).

Unsurprisingly, I ran out of time to do anything with this before we set out, and didn't find much more en route, beyond determining that the plug on the radio was finicky and it was too hard to convince the adapter's cable to mate with it. I bought a breakout cable for it–grumbling–and then took advantage of an electronically inclined friend's hack night to connect the parts together.

Pretty much everything went wrong from that point onwards.

The Bluetooth adapter was entirely happy to relay GPS data as long as somebody else connected to it first. But that's not the way that GPS mice (and apps that mimic them) work: a Bluetooth client must connect to them before they'll respond with GPS data. Trying to convince the adapter to initiate a connection was a multi-hour journey through inconsistent datasheets and discordant firmwares (firms-ware?); frustrating despite being a well-trodden path. The closest match–at least, the one whose sample output matched–was mostly untranslated and then the module didn't work as documented anyway, though I suspect here that it may have been attempting to connect to a particular UUID/port, and I still have no idea how to mess with Bluetooth at that level on a Mac.

Eventually I gave up, dug up somebody else's GPS mouse Android source, and talked it into making outbound connections. Then it became clear that the VX-8DR has NFI how to parse standards-compliant GPS output and demands fixed-width formatting because lazy firmware programmers. Also, the code was so reminiscent of my own approach to hacking together minimally functional Android apps that I spent approximately equal time feeling faintly nostalgic and wrestling with the urge to rewrite imprudently large chunks of it and send pull requests.

But with all that done and working, it was finally time to go for a drive! At which point I discovered that my Nexus 4 can't push data down an RFCOMM link without destroying the audio quality of the podcasts it's streaming through the car's speakers. Argh. Then it also became clear that the radio largely failed to punch a signal through the car's skin anyway.

So, back to the drawing board. On the upside, I rediscovered a lovely parkable perch for admiring the city lights from above.

dichro: (Default)
I find myself reading weev's lawyer's brief from earlier this year. It shows just how much I don't understand about the law. I wind up reasoning about these things by analogy, whereas I guess the ruling standard is to find matching precedent instead.

Argument 1, p29, argues that there was nothing wrong with collecting the email addresses because they were on a public website, not protected by a password; and merely using a public resource "in a way that the owners find dissatisfying" is not actionable.

I think the usual counter-argument to this is that your legal protection from theft is not predicated on locking your front door - even if someone finds your door is unlocked, they aren't legally authorized to take all your stuff on that basis.

(Sidebar: another argument might be that it isn't "theft" because the other party wasn't deprived of anything. Adapting the above analogy, if someone finds your door unlocked, enters your house, copies your bank account number/SSN/driver's license/etc, and passes those on to someone else who uses them for identity theft purposes, that's still not ok, right?)

But I don't think those are even all that relevant, since this seems to hinge on the question of what a password is. Sure, there's no interstitial screen with two separate text boxes, one for username and one for password (the latter only displaying stars when you type in it), before yielding the email address. But just hitting the bare URL gives you nothing. You have to supply an ICCID to learn an email address, and I can imagine an argument being made that the ICCID is, effectively, a password, since it's a secret only shared between the user of the iPad (by proxy) and AT&T.

"But wait", I hear you cry, "they issued them sequentially! Nobody generates password like that!"

Oh, grasshopper. If only that were the case. But even it were, again, it doesn't seem that the law should require that you issue good passwords to have legal protection. It should still apply even if the scheme is one an idiot would use for their luggage (but leave whoever came up with the scheme open to charges of neglect or incompetence or something).

There's a number of scraper-related cases cited as precedent, but they all appear to be following links from public information to (de facto) public information. I don't know, but I'm going to guess that the list of ICCIDs is not publicly available, and being able to generate one from another doesn't seem like the thing that would add you to the class of people authorized to access that information.

Again, analogies: if you happen to catch sight of a key, you could make a copy of it. But, for the purposes of legal access, it doesn't seem that that should put you into the same category as those who have received a physical key from an authorized person, no matter how easy the process might be.

The brief acknowledges that "[g]uessing someone else's password to gain access to another person's private account without permission constitutes a criminal act", so I guess their argument is that because the URL parameter that they were guessing was named "ICCID" rather than "password" makes it ok. I don't know if I find that enormously convincing.

Argument 2, p43, rings pretty true to me. There's one defense in there based on no password being guessed, which, per above, seems like splitting semantic hairs, but the rest of it seems reasonable, with one exception.

The second claim of part B of argument 2 states that New Jersey had no jurisdiction because the defendants, computers and information disclosure were outside the state. Somewhere earlier, I think, they alluded to NJ residents' email addresses being revealed as the jurisdictional basis.

On its own, that seems reasonable - if the defendants, computers and disclosure were entirely outside the country, would that mean no action could be taken by the US? Evidence suggests otherwise. So, if there were harm or the risk of harm down the road to NJ residents, it doesn't seem crazy that they'd have an interest in and argument for prosecuting there. It's probably not the best jurisdiction for it, but does that make it an invalid one? Dunno. Law is hard.

Argument 3, p49, seems solid (modulo, again, the password thing). It seems pretty obvious that there was no intent to commit identity theft, so it's hard to see why identity theft statues would apply.

Argument 4, p55, is a slam dunk. It's an extension of the jurisdictional points in argument 2, but cites law to support that if a jurisdiction isn't the best one (paraphrasing), it is indeed an invalid one.

Argument 5, p62, cites some pretty convincing precedent in support of its argument that the costs to AT&T should not have been considered. The costs were almost entirely due to a snail-mailed notice to customers that their information had been stolen, and there's a citation from "at least one district court" saying "determining and complying with customer security breach notification obligations" "do not qualify as loss under the CFAA."

It was a pretty interesting read. I'd been largely assuming that this was an Al Capone situation, where weev's proudly self-declared assholery came home to roost under flimsy legal pretenses, but now I'm not at all sure. The whole password analogy seems indefensible, and I'd expect him to get nailed for it, except that the jurisdictional challenge also seems equally indefensible. O'course, my technically-biased viewpoint on this means pretty much nothing to the system; I'm off to try to find what the legal minds filed or decided in response to this.

Skybergs

Nov. 20th, 2013 03:36 am
dichro: (Default)
I haven't given up on the whole floating thing. The notion that's been rolling around my mind of late is aerogel-filled balloons. It isn't a new idea, and even SEAgel is easily floatable at STP.

But the more interesting compound is the recently synthesized aerographene, which not only boasts an absurdly low density, but remarkably high compressive strength.

One of the things mentioned by smarter balloon people than I was the magnitude of the thermal stresses caused by solar heating. You have an envelope floating around, filled with gas, when the sun comes up. In short order, the sun heats the envelope and gas, raising the pressure inside significantly. Your envelope either bursts or vents. If it vents, then when the sun subsequently goes down again, envelope temperature will return to equilibrium, and the envelope will shrink, causing its lifting power to be dramatically reduced, thence likely plummeting to earth, dragged down by the weight of your payload.

If the envelope is filled with aerogel, two things come into play. One, the aerogel resists being compressed, meaning that the envelope will be at a slight negative pressure to the outside air, boosting buoyancy. The aerographene paper is irritatingly vague on compressive modulus, but it does cite two useful facts: compressive modulus is 8 kPa at a density of 5.1 mg cm-3 and varies with the square of the density. Since the paper claims that the lowest synthesizable density is 0.16 mg cm-3, one could handwave, but it's too hard to run up a spreadsheet at 3am to verify my suspicion that that's too small an effect to make much difference. I think the density of aerogel required to make a meaningful difference at any particular pressure significantly outweighs the air it's saving.

The other potentially useful property is insulative. These payloads all have electronics inside them, and electronics generate heat. Nesting the electronics inside an aerogel balloon might result in a high enough temperature to make a meaningful difference. 3am is once again preventing me from contemplating Stefan-Boltzmann's law and the diurnal stratospheric temperature curve with the focus those calculations require, insomnia notwithstanding.

There's a few other interesting points in the aerographene paper. The synthesis doesn't appear enormously complex to my ignorant eyes (although the synthesis of giant graphene oxide sheets, a precursor, doesn't fill me with enthusiasm). The raw materials aren't enormously expensive for experimental quantities. Lastly, the paper claims that production can be easily scaled to volumes of m3.

I dream of people using their swimming pools to grow aerographene skybergs, attaching a lawn chair, and floating away.
dichro: (Default)
There's a backlog–as always–to write about, what with the move back to the west coast, but one subject that has come up with surprising and gratifying frequency in my recent visits there is pronoun preference. Since I exacerbated the problem during my move east, people have often tried to guess, corrected themselves, followed each others' lead, desperately groped for syntax, and generally caused me more than a little amusement.

But for all that it's a little late for National Coming Out Day, let me take this opportunity to set the record straight, or at least–pace Adams–firmly crooked.

I should first confess that my own views on this have evolved in the last few years, and I'm in no way guaranteeing that they're now stable. But since I'm feeling pretty comfortable in my skin and soul for the first time ever, there's a reasonable chance that I won't wander too much further afield.

An ever-increasing number of people ask "which pronoun do you prefer?" It's always a good question to hear. It's someone saying that they recognize that this might be an issue that's important to you, and that they want to be sensitive to it and act in a supportive way. But despite appreciating the question and wanting to validate their concern, I've never had an answer to it that I felt particularly comfortable with.

Initially I devised a tortuous response that amounted to it not mattering when speaking to me directly, but when speaking about me, in the third person, they should choose whichever pronoun would make the story more interesting. In retrospect, I doubt that answer really helped anybody. I then came up with a worse one, which was to point out that here was an opportunity for my interlocutor to signal their gender reactionism or open-mindedness via their choice of gender pronoun to use for me, and that I'd then judge them in turn for it. It's amazing I've made any friends at all, ever, really.

Part of the problem with the question–aside from my apparently having an otherwise well-managed streak of asshole in my makeup–is the implicit gender binary. Which pronoun do I prefer of the two? Well... that's complicated.

Male pronouns don't fit too well for dressy occasions; female ones are similarly awkward when I'm ruthlessly exploiting body privilege to be lazy in the heat. But, most of the rest of the time, either works. Some of my friends consistently use one or the other, some switch, a tiny minority use nouveau constructs, if only in written form.

When required to identify on paperwork or online signups, I take the first option that's available and contextually reasonable from the list below. I'm OK with the first three; the remainder are compromises.
  1. gender-queer/-fluid/-etc
  2. other
  3. decline to state/prefer not to say
  4. transgender
  5. female
  6. male
I have two passports and two driver's licenses, from three countries. Of these four legal, official documents of identity, no two make the same statement about my gender. That works for me.

(For the record, it isn't because I particularly like having discordant ID; the four different systems that generate the four documents simply allow different sets of options from the list above. I'd be entirely happy for all my IDs to conform to my Australian passport, despite it confusing the heck out of United's check-in machines.)

Pronouns are hard. I have gotten them wrong with other people in ways that I am ashamed of, and doubtless will again. If I automatically make a mental classification that turns out to be incorrect, I often find it incredibly hard to switch to their preferred pronoun. I feel awful every time I get it wrong; for some of these people, I've run out of time to make it up to them. I have no desire to put my friends through any of this.

So here's the deal. My preference is to avoid gendered pronouns. Use my name. Use they/their. But if you can't, or get it wrong, don't sweat it overmuch; you won't ruin my mood or my day.

...but I will judge you.

1000 miles

Oct. 11th, 2013 10:04 am
dichro: (Default)
With a thousand miles on the clock and nearly that far into the roadtrip, I'm firmly in love with my first car. Not unlike the iTriumph (or me, I suppose), she's sexy, but hilariously impractical.

Possibly my favourite thing so far is the headlights. Aside from all the safety and visibility benefits, when you first switch the car on in the dark they go through a self-test routine that sounds for all the world like a compact camera's lens unfolding, and the light cones they project widen, narrow, move around, then focus on the ground; for all the world like she has woken up, stretched, studied her surrounds and is now standing ready with demurely downcast eyes.

The highlight of the trip so far has been thundering along mountain highways late at night in North Carolina, with a giant quarter-moon low on the horizon, and the lights automatically refocusing further away and all around (still with that delightfully cybernetic sound) whenever the road ahead is empty.

We spent the first three days of the roadtrip (all of them, to date) being rained on. The first time I opened the trunk, all the dewed rain cascaded from it onto the back window and rolled down towards the trunk. I was expecting it to flood in and drench our bags, but instead almost all of it was caught by the lip of the trunk and guided towards the corners, where it did fall into the trunk–and was neatly caught by a pair of perfectly shaped funnels that are so unobtrusive that I'd completely failed to notice them previously. Genius.

A large part of my excuse for buying a convertible was the lack of visibility in most cars I've driven compared to bikes. It's an imperfect solution: although I love how much I can see with a quick look over my shoulder when the roof down, the A pillars and upper windscreen surrounds are thick and block exactly the places one wants to see when stopped at the lights.

Worse, the rear-view mirror assembly is enormous (not least because it contains the forward-facing camera that monitors oncoming traffic to enable a lot of the automatic headlight smarts) and sits exactly between my eyes and any traffic lunging into the intersection from the right. 

With the roof up, as it has been for the entire trip so far, one changes lanes on faith to some extent. It's tempting to apply the bike approach here–if you're the fastest thing on the road, you don't need to check for overtakers–but agile as she is, that is unlikely to work out as well.

The main practicality grump is storage. There's none, basically, in the cabin. One can stash a wallet in the doors or the center console but little more. My attempts to mount a phone cradle to the sexy dashboard curves have so far failed, and we have an awful mess of cables powering phones and other devices perched precariously on laps, wedged under legs or in cup holders.

Her manual is 350 pages. I haven't made it very far in. She's a complicated wee beastie.
dichro: (Default)
Six years ago, on the cusp of my moving to the US, my grandmother suffered four heart attacks in rapid succession. Despite the emergency dash back to Perth from Sydney to see her, the emotional trauma, work stress, my telling her I loved her for possibly the first time in decades and so on, I can't find anything that I wrote on the subject. Livejournal is bare. I'm more than a little perplexed; how many things have fallen away because I didn't write about them?

I'm now on my way back from Perth again, to an even further home, in much the same circumstance: she had a stroke a week and a half ago, and has been more or less comatose in hospital ever since. I remember it seemed dire then as well, knowing that the next one could kill her, but it's worse now, since it seems that she's already gone.

We cling to the notion that on those rare occasions that she opens her eyes - usually in response to cleaning, turning, bathing or other insult by the nurses - she can see us, recognize us, and that it means something. We imagine a smile on her lips, a weak signal overlaid on the regular curve of her breathing. One of her own daughters had a stroke in her 70s, lay unconscious for three weeks, and recovered. But my grandmother is 102, and the medical verdict is for palliative care only.

There's a lot of decisions that are made away from the family and the patient, but it seems that they largely stem from a single, important difference: whether to treat or not. Rationally, statistically, I imagine that it must seem an open and shut case: patient, elderly, really elderly, major cerebral infarct, unresponsive, not improving for several days. Initially withhold IV nutrition, later remove oxygen and IV hydration as well. Let nature take its course.

I do wonder if there's some shadow of the public-vs-private healthcare debate here; if a private hospital would be content to keep a body in bed as long as somebody was willing to keep paying the bills. But in a public hospital, where beds are always in short supply, one triages even in the geriatric ward.

The medical specialists didn't always seem to be in agreement. RPH is a teaching hospital; perhaps it was one of the younger students who reacted with such shock a few days earlier when we first suggested taking her home, stating emphatically that while she might not get better in hospital, she definitely wouldn't at home. But the palliative care specialist, despite some epic circumlocutions, did eventually confirm that the IV hydration had been withdrawn to avoid prolonging life unnecessarily.

After I spoke to her privately, using the phrase "terminal congestion" in a leading question to establish my faux fides, she confided that a patient who reaches such an age in otherwise reasonable health can be assumed to have some biology working for her, and it's conceivable that with minimal maintenance she could linger for weeks, perhaps months; clearly unthinkable.

The family have decided to take her home. It'll be happening today, perhaps tomorrow. Palliative care are familiar with the scenario and can assist with all the necessary people, equipment and procedures. I won't be there; I'm taking my persistent wracking cough back home to meet other obligations. I tried to say goodbye; we're all hoping or pretending that it had some meaning, that there was enough of her there, locked away behind her untracking eyes to understand it. I do think she smiled.
dichro: (Default)
I'm a long way behind on posts, but luckily that lets me tie up two threads together that'd work poorly on their own. It's all about the games.

...in which I go to Sweden for space and Norway for therapy. )
dichro: (Default)
A couple of years back I did the Inka trail. Received wisdom had it that the climb up is not so hard, but it's the descent from the highest point (charmingly dubbed Dead Woman's Pass) that is truly murder on the knees. It was exactly so, and I was in pain thenceforth. A year or two after I returned it dawned on me that my knees were still hurting, and perhaps that was longer that it strictly needed to be. One x-ray later, the bad news: arthritis.

It turns out the word itself just means joint inflammation. The scarier versions are the systemic conditions like rheumatoid arthritis, but much of it - as in this case - is damage caused by mechanical trauma that doesn't portend anything particularly scary. The condition never gets better, but it can be kept stable by strengthening the muscles around the joint such that they brace the body's weight, rather than letting it rest on the abraded interface between the bones. Or something like that; I'm no doctor.

I'd been in the habit of walking everywhere in New York (modulo that unfortunate episode in Brooklyn), and doubtless that had exacerbated the problem somewhat, but there was an easy solution: I bought a folding bike. It's not an ideal solution for transport, since I'm terrified of leaving it locked up in the street, but it does the job for the daily commute, and it should strengthen exactly the right muscles in my legs to protect my knees without any impact to aggravate the damage, not to mention banking all those purported benefits of even mild regular exercise.

That does, in turn, cause some new issues. It had taken me a while to get the hang of podcasts, but I'm now a convert, NPR addict, and Radiolab supporter. I had about the right number of daily updating newscasts to cover my commuting time, but riding in this city, headphones would be a rather bad idea, which leaves me with needing to find the better part of an hour a day to keep up on the news. This is a problem, since much as I want to keep up on said news, audio is my least favourite way of doing so. If it's an option, I'd much rather read than listen. It worked on the commute since reading was infeasible, but forsaking reading time for podcast time ain't gonna happen. Somehow I need to find not only a heretofore underutilized daily window of time, but also one that I can't use for reading.

To spare you the suspense: I found one. It lines up with a problem that I'd occasionally fret over: my tendency to burn anything up to an hour in a light doze every morning, in that twilight consciousness between waking up and actually leaving bed. It's often a combination of poor sleep and laziness that varies widely and unpredictably day to day. Nonetheless, that's the time that I decided to repurpose to audio appreciation.

The problem is, of course, the variability. Triggering podcasts from an alarm wouldn't work; my time to sleep varies widely, and I'm careful to keep my work schedule tolerant of waking times anywhere from 7am to noon. Anything that cut short useful sleep would be wildly counter-productive - the older I get, the more I notice the stark difference in my productivity between a well-slept night and any other kind. So I need some way to measure how much sleep I've had in a given night, account for any accrued sleep debt, and pounce on any sign of wakefulness once I've slept as much as I need to.

Tricky.

Back at the first Quantified Self conference I'd caught the sleep-tracking bug and ordered a Zeo. It didn't yield any terribly impressive results, but there was a way to extract real-time data from it with a modicum of hacking that I found intriguing but never quite got around to playing with. Nonetheless, I had all the required pieces and I set to, hampered only by Zeo's rather poorly timed corporate collapse that probably correlated with the seeming unavailability of the experimental firmware that provided the real-time data. A quick appeal (Graph Search: "friends who like Zeo") solved the problem: the fig leaf disclaimer didn't block direct access to the download. Shortly thereafter: real-time EEG data FROM MY BRAIN.

I don't necessarily have a high opinion of the Zeo's analysis capabilities, but the real-time firmware comes with a data-processing library that decodes the raw signal from the sensor, FFTs it into a relevant frequency bins, and takes a guess every 30 seconds as to the brain state that is being monitored.

Messing around with the data led to the first approximation: watch just the gamma band (30-50Hz in this case, supposedly correlates with more complicated cross-hemisphere activity that tends not to happen during sleep) and measure the length of continuous runs on either side of a magic value. I'd much rather some kind of bimodal estimator that deduces the magic value for itself, but this seems to work well enough for the moment.

Then it's just a case of taking action. I haven't quite got the podcast thing working yet, since Doggcatcher's cloud sync stuff is stuck in restricted beta for the moment, but I settled for turning on my lights and starting up Spotify through my receiver. It may be a little too conservative at the moment, in that I think it could comfortably kick in fifteen minutes earlier, but then time sense when dozing is pretty unreliable, so some staring at the data is called for. Most importantly it has yet to wake me up too early.

Ultimately I should stop using the Zeo for this altogether. The corporate failure means replacing the headband may become difficult, and it needs to be done every three months or so. All of the consumable Zeo stock vanished from Amazon as I watched, only to be relisted at thrice the previous price by brand new sellers. But some have found that motion-based sleep analysis is comparable to the Zeo's data, so I should probably just point a webcam at the bed instead. What could possibly go wrong?
dichro: (Default)
I've been thinking a lot, of late, about the pieces of my life that are tied into Internet services.

I'm in the privileged position of being able to afford to spend money on things, but nonetheless I often balk at paying for online services. Getting a paid Github account was a surprisingly difficult step at the time. But on recent reflection, I'd much rather be paying for a service I like than putting up with a free one that I don't, and right now I'm particularly enamoured with the idea of financially supporting smaller, scrappier, single-product services for sympathy with their product philosophies.

Long(ish)-form writing online - blogging, I suppose - is one that I've been wringing my hands about for a while. I've tried some alternatives but haven't found anything as good as LiveJournal was years ago. I left because it was getting all ghost-towny, I think, and Twitter had supplanted it for many things, but I do miss the community. Most of whom are now on Facebook instead. But I still want somewhere to dump and store longer musings.

For some reason I'm reluctant to just go back to LJ. It probably has something to do with zee Russians, though I couldn't tell you why - they're no worse than NSLs, and I don't know that I should expect much difference from private business either. So here I am, at least for now.

I'm probably going to reactivate my Flickr account. Not only does it already have a heap of my photos on it, but I've also been pining for the API pretty much since the moment I defected to Picasa. My favourite thing (that I never used): replacing a photo without disrupting the metadata. I'm imagining uploading downsampled photos direct from phone and using resulting comments/favourites to prioritize replacing the low-res images with full-res version. Yahoo isn't exactly a single-product company, but I can't think of another property of theirs that I care about (or... at all, actually), so perhaps that's close enough.

Finding a replacement for Reader is obviously important, but I don't have one yet. Newsblur, maybe? My consumption of news through Reader is a little erratic, but certainly valuable to me, if perhaps not quite as much as for the more vocal cohort of recent weeks.

Google Voice is fairly critical for my contactability for official matters, but between number portability and how little I really use it, it's a meh issue. Perhaps if there was a service that did a better job of integrating voice comms into a unified inbox, but then I can't really imagine abandoning GMail anytime soon.

Mint is a bigger deal. I really like having all my transaction data consolidated in one place (and I'm sure they do too) but I've been increasingly uneasy about them since they were acquired by Intuit. I don't know how a smaller service would keep up with their collection of scrapers/interfaces, though - not that they cover all of my accounts as it stands, anyway.

I have a vast quantity of e-books from Amazon, who - Liz assures me - are Evil(tm). But the bug bit me hard from traveling with my first Kindle. Not having to worry about reading material before you're at the gate or even in your seat was great. But I've decided that I miss real books too much, miss seeing what people are reading on the subway, and am willing to deal with the relocation hassles, particularly for the kids. At least for the more worthwhile books, anyway. So now I have a steadily growing physical unread pile that I peer into fondly from time to time.

I already give money to Spotify, partly for the sake of having music to play at parties sans advertising, the rest to have my receiver play my Spotify playlists directly; and I'm entirely sanguine about not technically owning any music. I'm still not very good at making time to listen to it.

I pay for Google Drive too, though I do wish Dropbox would get their shit together. I'd much rather share credentials for their service (though only intentionally, natch). It's probably time to take storage seriously, though, which probably means doing something complicated with keeping the bulk of the data in AWS, a working set locally, and relying on something like Ceph to magic them together. I never want to have to care about a failing hard disk again, nor about building computers for home.

Authentication is also somewhat troubling. Logging into random sites with Facebook/Google/Twitter authentication is pretty great, particularly those that set up a complete user account with the details without any further prompting. At the same time, once you start reflecting on how much you're trusting them with... yeesh. I started using algorithmic passwords after losing an unencrypted laptop to burglary, so could just stick with per-site passwords, but the convenience is pretty great.

Which I guess comes back to Facebook. I'm rather intrigued by the possibilities of Graph Search, and wishing somewhat wistfully that I'd been checking in to places there instead of on Foursquare, and there's far too many people that I only maintain contact with through Facebook. I just discovered their free-ish business cards too, and immediately ordered some. They're going to be tough to leave.
Page generated Oct. 19th, 2017 04:23 pm
Powered by Dreamwidth Studios