Bela Lugosi's Dead, Jim

The Horse, The Wheel And Language

Fri, 04 Dec 2009 21:16:17 GMT

The Horse, The Wheel And Language, David W. Anthony, ISBN 978-0-691-05887-0

This is another look at the question of when and where Proto-Indo-European (the inferred common ancestor of many the modern and ancient languages found in areas ranging from Europe to India) was spoken, and why, how and when it spread. What makes it worth a new work on the subject is not just new research, but also the improved availability of archaeological research from eastern Europe (both from the Soviet era and later).

The author spends a couple of chapters discussing the history of the question and arguing that reconstruction of unattested languages is indeed possible before moving onto meatier questions such as timing, settling on 2500BCE as a latest terminal date for PIE, based on the youngest IE branches (which are also among the first to be written). Moreover he argues on the basis of the various elements of related vocabulary, that PIE cannot have started to separate into its daughter language at least until its speakers’ adoption of the wheel, and that the archeology the wheel places this somewhere between 4000 and 3500BCE (with the possible exception that the Anatolian languages separated earlier).

This dating also forms the core of his response to the hypothesis that the IE languages spread with farming: the dispersal date it requires is just too early.

Timing, farming vocabulary and relationships to other reconstructed proto-languages (such as Proto-Uralic) are then used to argue for a homeland on the steppe north of the Black Sea, the familiar territory of the Kurgan Hypothesis. He proceeds to lay out one possible sequence of events, backed by considerable argument. This forms the meat of the book (see below for a summary).

The book is well supplied with maps and diagrams showing archaeological finds. Less diverting are the table of radiocarbon dates associated with various sites, though fortunately for anyone interested in when things are likely to have happened, dates are well represented in the text making these inessential. As well as historical-linguistic and archaeological research the author draws on evidence of past climate change (to offer reasons for some of the archaeologically observed changes) and documents his own extensive research into the the wear patterns made on horses’ teeth by their bits, although unfortunately a shortage of archaeological finds so far makes this less directly relevant to the question of when horses were first ridden.

The relationship of the steppes to the settled societies of the Middle East is also explored. The Indo-European languages didn’t develop in a vacuum, isolated from other developments in the world, but parallel to them and sometimes in (often indirect) interaction with them.

An interesting diversion concerns the current and ancient cognates of the much-abused Aryan. The word that speakers of proto-Indo-Iranian used to describe themselves was probably *arya-, from PIE *h₄erós “tribe”. The best known cognates apart form the modern word itself is probably country name Iran. Interestingly derivatives of the word appear in non-Indo-European languages (whose ancestral languages were presumably therefore spoken in proximity to Indo-Iranian), with Finno-Ugric *orya, Pre-Saami *oarji “southwest” and ārjel “southerner”, Finnish orja “slave” and Estonian ori “slave”.

This book is well worth reading as an answer to the Indo-European homeland question.

Anthony’s Answer

(Errors in this extremely compressed summary are of course mine!)

Farming (of cattle, sheep and grain) the European edge of the steppes, then occupied by foragers, around 5800BCE. Anthony argues, with analogy to modern examples, that the archaeological frontier here is persistent enough that it may have been a linguistic frontier too. Between 5200 and 5000BCE farming was adopted by some of those foragers; the new way of life, and probably the dialects ancestral to PIE, spread over the steppe. Horseback riding probably originated on the steppe around 4200BCE. (But riding horses into battle was much later, so PIE speakers weren’t early analogs of the Huns or Mongols.) The greater range of mounted herders led to expansion and conflict, and a cooling of the climate increased the need for movement. One result was the expansion of (archaic) PIE into southeastern Europe, giving rise to the Anatolian languages (e.g. Hittite, the earliest attested Indo-European language). Around 3500BCE some of the easternmost speakers of PIE migrated thousands of kilometers eastwards, producing the Afanasievo culture and thought to have originated the (now dead) Tocharian languages.

The wheel reached the steppe 3500-3300BCE, producing the the Yamnaya culture, which used wagons to maintain a predominantly mobile herding way of life: these people are identified with the speakers of late PIE. A number of factors are identified for the spread of IE languages westwards into Europe. Firstly IE speakers probably had more and better horses, allowing them to grow rich by trading them, to manage larger herds, and to raid more effectively. Secondly, inferred properties of Yamnaya society are suggested as easing the integration other groups into Yamnaya communities, leading to new speakers adopting the language. Three groups of languages find their genesis here: Italic (e.g. Latin and French), Celtic (e.g. Welsh) and Germanic (e.g. English). While migration may well have been involved in some cases the author emphasises that there was no “Indo-European invasion of Europe”.

From 2100BCE fortified towns of the Sintashta culture appeared east of the Urals. These people had (and perhaps invented) chariots, and their graves show many similarities with the rituals described in the ancient Sanskrit text of the Rig Veda, and they probably spoke the ancestor of the Indo-Iranian languages (e.g. Persian, Sanskrit and Hindi).

How big a stocking?

Fri, 04 Dec 2009 18:39:38 GMT

I'm actually more interested in how I keep them out.

Flashforward 8-10 (spoilers)

Mon, 30 Nov 2009 22:02:19 GMT

Episode 8

Where were we? The blue hand guys were busted but we’ve not yet found out any more about the hits on the primary cast. Noh’s finally owned up and Al Gough killed himself to ensure his vision doesn’t come true. Aaron Stark’s daughter Tracy appears to have turned up, looking pretty well for someone who’s supposedly dead.

The newspaper headline of “The Future Can Be Changed” is a good way of emphasizing what’s on peoples’ minds. It seems to have put Mark and Olivia back on track with each other.

Names on Lloyd Simcoe’s email: ...no, didn’t get them. Sorry l-) Myhill is probably one of them (from Simon and Lloyd’s conversation). Simon’s analysis that causing mass death qualifies one for godhood suggests a rather grim pantheon! The emphasis of Simcoe’s card tricks did make me think that his prestidigitation was going to turn up later though I’d not expected it to be in the same episode.

“I called 911 but they put me on hold” - is that remotely likely? There are some reports suggesting that response time for emergency calls made from cellphones can indeed be pretty poor. Parovsky/Perowski/???’s activities are surely involved in the flashforward somehow.

I see that the NSA have magical image-enhancement technology, normally it seems to be available to local police forces too...

The rings delivered by the three stars guy have an α on; suspect zero was probably wearing one of them. Did he willingly transmit them to Parovsky? Perhaps the three stars guys are form the Jericho PMC? If so then the village Jericho took out almost certainly had something to do with the bigger picture. (Benford’s character thinking that there could only be one person with a three star tattoo is rather ridiculous.)

Episode 9

So Bryce was suicidal because he has cancer. I do believe actually that’s a (admittedly rather minor) loose end tied up. Presumably the victim of his, er, car-park rage saying he was “dead” was supposed to be ironic. It looks rather like he and Keiko meet in the restaurant because ... they seems themselves meeting in the restaurant, though.

Aaron’s slow-burning alcohol disagreement with his daughter was much better done than Noh leaning on Levy for a copy of his own telephone conversation.

It’s a pleasing symmetry between Keiko’s and Bryce’s quests to find each other. Bit of a blunder not to spot each other on the same plane though.

Episode 10

Interesting that they mostly refer Nhadra Udaya as “Persian” and only occasionally as “Iranian”. To be honest I don’t know what the preference among Iranians is, though I’d certainly picked up the impression that “Iranian” was the usual name used by Westerners. Didn’t quite get the Iranian lady’s name, anyway.

Simon Campos’ denial of NLAP’s involvement contradicts his opening gambit to Dylan several episodes back that they were responsible. It was good to see him taken down a peg, anyway. Given this is a time travel story my guess would be that someone used his design in their time machine, but independent repreinvention would be another possibility.

Simcoe’s introduction of the Many Worlds Hypothesis is presumably meant to make us suspect that the flashforwards may be representations of alternative futures: something that could explain how Gough generated an inconsistency (recalling that Noh’s inconsistency with his fiancée has now been undermined by her actually paying attention to her vision).

Noh being shot with Benford’s gun, assuming pace the above that it’s safe to trust to the reliability of the visions, has two possible explanations at this point: one is that someone else did it, Benford having handed over his gun, and the other being that Noh is the mole and Benford discovers this (having recovered his FBI status by at some later point in the plot).

Benford’s gun’s serial number is A561984. Noh’s badge number is 4587.

Characters

Name	Actor	Description	Notes
Mark Benford	Joseph Fiennes	FBI agent	Husband of Olivia Benford Father of Charlie Benford Colleague of Demetri Noh
Aaron Stark	Brian F. O'Byrne	AA-goer	Confidant of Mark Benford
Tracy Stark		Soldier	Was thought to be dead
Nicole Kirby	Peyton List	Student	Babysitter for the Benfords Japanese speaker, helps out Bryce
Olivia Benford	Sonya Walger	Doctor	Wife of Mark Benford Mother of Charlie Benford Apparent future lover of Lloyd Simcoe Colleague of Bryce Varley Operated on Dylan Simcoe
Demetri Noh	John Cho	Junior FBI agent, apparently doomed	Colleague of Mark Benford Fiancé of Zoey Andata
Zoey Andata		Gabrielle Union	Fiancée of Demetri Noh
Bryce Varley	Zachary Knighton	Formerly-suicidal doctor	Colleague of Olivia Benford
Lloyd Simcoe	Jack Davenport	(Self-described) annoying parent	Father of Dylan Simcoe Apparent future lover of Olivia Benford Apparently unwilling co-conspirator of Simon
Dylan Simcoe	Ryan Wynott	Injured autistic child	Son of Lloyd Simcoe Saved by Olivia Benford
Charlie Benford	Lennon Wynn	Winsome child	Daughter of Mark and Olivia Benford Saw D. Gibbons in vision and didn't like him
Stanford Wedeck	Courtney B. Vance	Assistant Director FBI LA Field Office	Boss of Mark Sanford and Demetri Noh Former fixer for the current president
Joyce Clemente	?	US Senator	Political opponent of Stanford Wedeck
Janis Hawk	Christine Woods	FBI agent	Girlfriend of Maya Pregnant in her flashforward
Al Gough	Lee Thompson Young	FBI agent	Would have killed Celia had he lived, but suicided
Celia ?	?	Mother	Would have been killed by Al Gough
Keiko Arahida	Yûko Takeuchi	Robot researcher at Nakahara	Seen by Bryce in his vision?
Maya			Girlfriend of Janis Hawk
Alda Hertzog	Rachel Roberts	Terrorist
Nhadra Udaya	Shohreh Aghdashloo	Spook	Sheltering D. Gibbons Learned of Demetri Noh's death
D. Gibbons	Michael Massee	Obscure hacker/bomber	Telephone interlocutor of suspect 0 Killer of Pigeon local cop
Suspect 0		Awake during the flashforward	Telephone interlocutor of D. Gibbons
Simon Campos	Dominic Monaghan	Sinister guy on phone	Tells Lloyd Simcoe they're responsible for the “greatest disaster in human history” (presumably the flashforward). Uses expertise in QM to pick up women on trains.
Old-looking bloke		Ring collector	Has six out of seven α rings
“Reynaud”	Jeff Slingerland	Blue Hand organizer

What I did with my time off

Fri, 27 Nov 2009 22:03:42 GMT

Got a roofer to look at our loft, which has a squirrel problem. We need our eave combs (i.e. the things that block the gap at the bottom of the tiles) replacing. Quote expected next week.
Taught DisOrder how to use Secret Rabbit Code.
Added drag and drop and playlist support to Disobedience.
Wrote a new manual for Disobedience. It's also grown some help buttons which link into the appropriate bits of the manual.

Drag And Drop The Hard Way

Wed, 18 Nov 2009 11:26:16 GMT

Disobedience, DisOrder's GUI client, uses a GtkTreeView to display its current queue of tracks (and for various other purposes). It allows rearrangement of the queue by drag and drop. The high-level support for this in GtkTreeView is rather unsatisfactory for two reasons:

You can only drag one track (row) at a time. It should be possible to drag the whole selection.
You cannot drag between widgets. Playlist editing in particular is likely to make this important.

(Actually the model where it reports the effects of the drag and drop by inserted/deleted signals on the underlying model is also quite inconvenient for me: Disobedience has to translate it back into a move which it communicates to the server.)

Fixing this involves doing rather a lot of the drag-and-drop work yourself, and while the documentation does exist it is not always clear or even complete. There is existing code out there, and I referred to it while working some of this out; but describing it in English online seems worthwhile. In addition, while some of the code is inextricably mixed with Disobedience's internal workings, a couple of independent modules are presented, which should be usable in other C programs.

This articles assumes some familiarity with GTK+. I have tried as far as reasonably possible to remove DisOrder-specific issues from the discussion here, but note that this isn't always possible.

1. Low-level drag and drop

See queue-generic.c for the full code of this section in context.

1.1 Setup

First it's necessary to define possible types of data to drag using a target array.

static const GtkTargetEntry queuelike_targets[] = {
  {
    (char *)"text/x-disorder-queued-tracks",
    GTK_TARGET_SAME_WIDGET,
    0
  },
  {
    (char *)"text/x-disorder-playable-tracks",
    GTK_TARGET_SAME_APP|GTK_TARGET_OTHER_WIDGET,
    1
  },
};

The first field is a string drag type; if there's a standard way to choose these I've not yet discovered it, but MIME content types seem to be popular.

The second is a set of flags defining possible relationships between source and destination. For queued tracks, dragging is restricted to rearrangement within the same widget. For tracks to play, it is restricted to other widgets within the same application.

The final field is an integer index value used to identify the type. It will be supplied to some of the signal handlers. Note that it does not identify the particular entry in the targets array, so having two entries with the same type name but different indexes won't work very well!

The name queuelike_targets: a queuelike is a list of tracks: instances are the queue itself, the recently-played list, the recently-added list, and in the future playlists. They correspond to tabs in the UI. All of them will be updated by the server, but the queue can also be updated by the user, by adding, removing and rearranging tracks. The goal, of course, is for rearrangement to be possible using drag and drop within the queue and for adding to be possible by dragging from outside it.

It is necessary to mark the widget as both a drag source and destination. For queuelikes this is done as follows:

gtk_drag_source_set(ql->view,
                    GDK_BUTTON1_MASK,
                    queuelike_targets,
                    sizeof queuelike_targets / sizeof *queuelike_targets,
                    GDK_ACTION_MOVE);
gtk_drag_dest_set(ql->view,
                  GTK_DEST_DEFAULT_HIGHLIGHT|GTK_DEST_DEFAULT_DROP,
                  queuelike_targets,
                  sizeof queuelike_targets / sizeof *queuelike_targets,
                  GDK_ACTION_MOVE|GDK_ACTION_COPY);

(ql->view is just the GtkTreeView used to display whichever queuelike it is.)

GDK_ACTION_MOVE restricts the possible set of drag types (move, copy, etc). The queue requires arbitrary rearrangement, so it is capable of participating in moves both as a source and as a destination; and in files being dragged into it from outside, hence its ability to take part in copy-dragging as a destination (only).

GTK_DEST_DEFAULT_HIGHLIGHT arrange for whole-widget highlighting to be done automatically for destinations. I don't know why you wouldn't set this.

GTK_DEST_DEFAULT_DROP arranges for the "drag-drop" signal to be handled automatically by destinations. It will request the data from the source; the "drag-data-received" signal will be used to actually communicate the data to the destination. There is a caveat about this: if you ever request the drag data in any other way then you can't use this option as it won't be able to tell the difference between your drag data and its.

A variety of signals are then handled. For rearrangeable queues, we handle "drag-motion", "drag-leave", "drag-data-get" and "drag-data-received". These are discussed below.

For non-modifable queuelikes the code is a subset of the above: the gtk_drag_source_set() call (but with GDK_ACTION_COPY), and the "drag-data-get" signal handler. The widget is not marked as a drag destination and the other signals are not handled.

1.2 "drag-motion" handler

This signal applies to the destination and is emitted whenever the pointer moves over it during a drag. It has two main jobs: decide whether a drop at the current location is appropriate, and provide visual feedback about this.

These two operations are fairly separate in Disobedience though in other programs might well be more entangled. For the first half, the logic used by GtkTreeView is replicated, although in a simpler fashion, taking advantage of the small set of actions that Disobedience uses:

if(dc->suggested_action) {
  if(dc->suggested_action & (GDK_ACTION_MOVE|GDK_ACTION_COPY))
    action = dc->suggested_action;
} else if(dc->actions & GDK_ACTION_MOVE)
  action = GDK_ACTION_MOVE;
else if(dc->actions & GDK_ACTION_COPY)
  action = GDK_ACTION_COPY;

The full version can be found in gtk_drag_dest_motion() in gtkdnd.c - it only runs if GTK_DEST_DEFAULT_MOTION is set, which is why Disobedience needs to reimplement it.

Having done that it is checked whether this widget is a suitable destination. This is where the GtkTargetEntry array defined above comes into play.

if(action) {
  if(gtk_drag_dest_find_target(w, dc, NULL) == GDK_NONE)
    action = 0;
}

This much is enough to call gdk_drag_status() to indicate whether the drag will succeed and if so what action to use. If this depended upon the content of the drag then a more sophisticated approach would be required: the "drag-motion" handler would have to use gtk_drag_get_data() and field the response with the "drag-data-received" handler. This in turn would mean supplying a "drag-drop" handler which reimplemented the logic of GTK_DEST_DEFAULT_DROP but was also capable of distinguishing the intermediate checking drop from a final drop.

gdk_drag_status(dc, action, time);

The second half is implemented by calling gtk_tree_view_get_dest_row_at_pos() to find the target row. In Disobedience's case there is a bit of extra complexity. Firstly, you can't drop onto a row, only between rows, so GTK_TREE_VIEW_DROP_INTO_OR_BEFORE is replaced with GTK_TREE_VIEW_DROP_BEFORE and similarly for the _AFTER variants.

gtk_tree_view_convert_widget_to_tree_coords(GTK_TREE_VIEW(w),
                                            wx, wy, &tx, &ty);
  if(gtk_tree_view_get_dest_row_at_pos(GTK_TREE_VIEW(w),
                                       wx, wy,
                                       &path,
                                       &pos)) {
  switch(pos) {
  case GTK_TREE_VIEW_DROP_INTO_OR_BEFORE:
    pos = GTK_TREE_VIEW_DROP_BEFORE;
    break;
  case GTK_TREE_VIEW_DROP_INTO_OR_AFTER:
    pos = GTK_TREE_VIEW_DROP_AFTER;
    break;
  default: break;
  }

Secondly, if the drop is above the first row or below the last row gtk_tree_view_get_dest_row_at_pos returns no path, so the result has to be adjusted accordingly. Note that to distinguish between the two cases it's necessary to use tree coordinates, not the widget coordinates that the callback is supplied.

} else if(gtk_tree_model_get_iter_first(model, iter)) {
  if(ty >= 0) {
    do {
      *last = *iter;
    } while(gtk_tree_model_iter_next(model, iter));
    pos = GTK_TREE_VIEW_DROP_AFTER;
    *iter = *last;
  } else {
    pos = GTK_TREE_VIEW_DROP_BEFORE;
  }
  path = gtk_tree_model_get_path(model, iter);
}

model is the underlying GtkTreeModel; in this case a GtkListStore. Remember that paths must be freed after use.

It's worth pointing out that gtk_tree_view_convert_widget_to_tree_coords() only appears in GTK+ 2.12. So this code won't work on versions older than a couple of years.

1.3 "drag-leave" handler

This signal applies to the destination and is very simple: the visual feedback showing where data will be dropped is removed.

gtk_tree_view_set_drag_dest_row(GTK_TREE_VIEW(w), NULL, 0);

1.4 "drag-data-get" handler

This signal applies to the source. Its job is to put together the data to be dropped. In Disobedience, this is always because the data is about to be dropped but in the general case it could be because some candidate destination needs to be able to do fine-grained rejection of possible drops. See the discussion of GTK_DEST_DEFAULT_DROP above.

In Disobedience's case, the data will be a string consisting of track IDs and names. Once it has been constructed it is communicated back to GTK+, which will subsequently make it available to the destination via the "drag-data-received" signal.

gtk_selection_data_set(data,
                       GDK_TARGET_STRING,
                       8, (guchar *)result->vec, result->nvec);

GDK_TARGET_STRING indicates that the data has string type and 8 is the number of bits per unit. This bit seems to be particularly opaquely documented, I got there with a fair bit of trial and error.

1.5 "drag-data-received" handler

This signal is called on the destination when droppable data has been received. (As discussed earlier) it can in principle happen any time during a drag and drop operation, but in Disobedience's case it only happens when the final drop occurs. The data is retrieved as follows:

result = (char *)gtk_selection_data_get_text(data);

To figure out where to drop, the same code as used by the "drag-motion" handler is called; it's important that these match as otherwise the visual feedback and the actual action will not be consistent.

2. Multiple-row drag and drop

GtkTreeView's built-in drag and drop support only handles one row at a time, which is not really adequate, and is one of the reasons for going to so much effort reinventing the wheel. The code above will cope with multiple rows just fine but there is one problem: as soon as the user presses a button to start a drag, the selection will be reduced to just one row!

The answer is found in Quodlibet's qltk/views.py: intercept the button press and release events and suppress the change to the selection unless they're in the same place.

My C implementation of this concept can be found in multidrag.c. The code is very liberally licensed.

The "button-press-event" handler does the following things:

Discard any cached click location and unblock selection.
If the click isn't button 1, or is modified (shifted etc), return.
If the click doesn't identify some path in the tree, return.
If the path is not selected, return.
Otherwise, the click must be an unmodified left click on an identifitable, selected path in the tree. Attempts to modify the selection are blocked, and the pointer position is saved.

Blocking the selection is done via gtk_tree_selection_set_select_function(). The pointer position is stashed using g_object_set_data() and g_object_get_data().

The "button-release-event" handler does the following:

If no click is remembered, return. In this case it's assumed that the selection can't be blocked, too.
Unblock selection attempts.
If the click is in the same place, find the path at that location and point the tree view cursor at it. This mimics the effect that the original click would have had.

This source file also handles the "drag-begin" signal and uses it to construct a drag icon out of the rows to be dragged. This isn't discussed in detail here.

3. Automatic scrolling

There is one more part of the jigsaw: when the pointer is near the top or bottom of a destination window, it should scroll up or down respectively, to allow drops in locations outside the current viewport. GtkTreeView only does this natively if you use the high-level interface and the code that implements it isn't exposed to callers.

I solved this by bringing this code out into autoscroll.c. The code is derived directly from the GTK+ source code (although modified to work without direct access to private data structures) and therefore retains the LGPL2+ license.

There are two entry points: autoscroll_add() enables automatic scrolling when the pointer is near the top or bottom of the window and autoscroll_remove() disables it.

autoscroll_add() is called at the end of the "drag-motion" handler, and autoscroll_remove() in the "drag-leave" handler. The GTK+ equivalents are called in a few other contexts but in fact "drag-leave" is raised when a drop completes as well as when the pointer leaves the destination, so this turns out to be sufficient.

References

GtkWidget. Some of the signals documented here are used.
GTK+ drag and drop functions. A number of these functions are used.
GDK drag and drop support. A few bits of this are needed.
GTK+ Tree And List Widget Overview and GtkTreeView. General documentation for the tree view and related widgets.
DisOrder.
Other posts about DisOrder.
Quodlibet.

Dr Who

Sun, 15 Nov 2009 20:06:32 GMT

I think someone has been watching Silent Running.

Flashforward 5-7 (spoilers)

Mon, 09 Nov 2009 22:29:12 GMT

Episode 5

Stanford blackmailed the president!?! That's not going to end well. The White House seems to have escaped physical damage but the Capitol's dome got broken (a cliché I think?) although the night shot towards the end didn't appear to show this. Could just have been the angle, but.

Bit insensitive to Mark for his colleagues to celebrate the end of the Washington trip in a bar, much as it may conveniently emphasize devil on his back.

Continued excellent visuals.

The gunfight at the end was a little OTT, and thought quite entertaining for it I can't help but feel that the bounding of the episode with a cliffhanger was a little forced. At a stretch the Clemente announcement could have done the job, failing that the hit on Janis Hawk (two attackers on one target) was arguably rather more plausible than the equally matched attack on Stanford Wedeck and his colleagues. (Just how realistic is using a rocket launcher in a confined space anyway?)

Episode 6

“Never go anywhere unless I'm going to get the real thing” seems like a good policy. And the QM/sex superposition is amusing l-)

The Benford/Simcoe revelation was done well enough as such but it was really there to bring out the conflict between the Benfords themselves; on the one hand it's a bit of an artificial push, on the other hand very little substantive has actually happened lately (gathering yet more obscure clues doesn't really pay off until they start to fit together).

Episode 7

Back to some striking visuals, especially around Al's suicide attempt (the pan across his face for instance), after a bit of a hiatus in the last episode.

The Blue Hand meeting has shades of Fight Club although their security protocol is a bit flawed! (But then, so was “You do not talk about Fight Club”) The question is, though: were they there to advance the plot, or just to explain Noh's moment of honesty?

We were promised a revelation: but it looks like the only revelation is that the visions aren't certainties, which we could already infer from the inconsistency between Demetri and Zoey's visions anyway.

Bryce and Nicole's kanji symbol is said to be for belief. It's well represented on tattoo websites l-) but more usefully can be found on Wiktionary 信 which suggests a Romanization of “shin” and a meaning of trust or faith.

Maurice Raynaud was a C19^th doctor who discovered Raynaud's Disease, which affects circulation in extremities, leading to blueness - an obvious connection l-)

Characters

Name	Actor	Description	Notes
Mark Benford	Joseph Fiennes	FBI agent	Husband of Olivia Benford Father of Charlie Benford Colleague of Demetri Noh
Aaron Stark	Brian F. O'Byrne	AA-goer	Confidant of Mark Benford
Nicole Kirby	Peyton List	Student	Babysitter for the Benfords Japanese speaker, helps out Bryce
Olivia Benford	Sonya Walger	Doctor	Wife of Mark Benford Mother of Charlie Benford Apparent future lover of Lloyd Simcoe Colleague of Bryce Varley Operated on Dylan Simcoe
Demetri Noh	John Cho	Junior FBI agent, apparently doomed	Colleague of Mark Benford Fiancé of Zoey ??
Zoey Andata		Gabrielle Union	Fiancée of Demetri Noh
Bryce Varley	Zachary Knighton	Formerly-suicidal doctor	Colleague of Olivia Benford
Lloyd Simcoe	Jack Davenport	(Self-described) annoying parent	Father of Dylan Simcoe Apparent future lover of Olivia Benford ?Unwilling co-conspirator of Simon
Dylan Simcoe	Ryan Wynott	Injured autistic child	Son of Lloyd Simcoe Saved by Olivia Benford
Charlie Benford	Lennon Wynn	Winsome child	Daughter of Mark and Olivia Benford Saw D. Gibbons in vision and didn't like him
Stanford Wedeck	Courtney B. Vance	Assistant Director FBI LA Field Office	Boss of Mark Sanford and Demetri Noh
Joyce Clemente	?	US Senator	Political opponent of Stanford Wedeck
Janis Hawk	Christine Woods	FBI agent	Girlfriend of Maya Pregnant in her flashforward
Al Gough	Lee Thompson Young	FBI agent	Would have killed Celia had he lived
Celia ?	?	Mother	Would have been killed by Al
Maya			Girlfriend of Janis Hawk
Alda Hertzog	Rachel Roberts	Terrorist
D. Gibbons		Obscure hacker/bomber	Telephone interlocutor of suspect 0 Killer of Pigeon local cop
Suspect 0		Awake during the flashforward	Telephone interlocutor of D. Gibbons
Simon	Dominic Monaghan	Sinister guy on phone	Tells Lloyd Simcoe they're responsible for the “greatest disaster in human history” (presumably the flashforward). Uses expertise in QM to pick up women on trains.
“Reynaud”	Jeff Slingerland	Blue Hand organizer

Ice, oh, one hundred!

Fri, 23 Oct 2009 20:06:52 GMT

View Poll: #1475390

Flashforward 1-4 (spoilers)

Mon, 19 Oct 2009 21:09:17 GMT

Episode 1

There's some nice camerawork in there, e.g. when Bryce Varley is considering his end.
A lot of the first episode must have practically written itself: the junior (of course) fed sees nothing, the senior one sees himself investigating the flashforward, his wife sees herself with another man ... the list goes on, ending of course with the one mysterious man who didn't pass out. The obvious big questions are articulated: how and why did this happen, is the future seen a certain one or merely a possible one?
The devastation after the flashforward seemed a little overdone. Chaos on the roads is plausible enough and one helicopter hitting a building is fair enough, but exactly how many helicopters were there actually in the air along the line of a single road? (And how did did keepers falling asleep for a couple of minutes cause zoo animals to escape? Thought the kangaroo is probably there to make you wonder if animals had the same experience and if not why not.)
The story is, at least initially, bounded: the date of the visions from the future is known, requiring the story to get there rather than rambling on indefinitely. Personally I'm inclined to see this as positive though of course,

Episode 2

More nice photography, e.g. the shot of the helicopters over suburbia from directly above.
Mark's boss's story is entertainingly horrible l-)
Varley is shaping up as the philosopher of the show:
“Everything's going to be OK, Mr Simcoe”
“How can you be so sure?” “I've seen the future.”
Mark is letting the vision guide the investigation. There's a real risk here that's he not analyzed, which is that he's heading down a path to the endpoint of the vision, not to the reality behind the flashforward.
“He who foresees calamities suffers them twice over” is attributed to one Beilby Porteus, an Anglican bishop and leading abolitionist.

Episode 3

Either Noh's fiancee is lying or at least one of their visions (and the phone call he got) is false or profoundly misleading. Does he have a twin brother perhaps? Aaron's vision seems to be inconsistent with the grave evidence too (assuming he didn't lie about it). Noh + girlfriend have inconsistent memories of how they got together too - perhaps the point is simply that future memories are inconsistent too!

According to Wikipedia, Kabbalah = הלבק = Kuf Bet Lamed He = which by the Gematria rules is 100+2+30+5 = 137. So Geyer's explanation is possible. A causal loop is inadequate justification for liberating Nazis, Benford needs to get a grip on himself.

There seems to be a giant neuralizer in southern Somalia. The Ganwar region may be fictitious (or at least the web hasn't heard of it!) Geyer's claim about the significance of a ring on the left thumb doesn't find any online support either.

Episode 4

Another reminder of the destruction when the flashforward happened; I wonder if there'd be really that much chaos?
Ned Ned is quite entertaining l-)
If the FBI can't get some help from the CIA... obviously they hire a computer hacker. I'm surprised they didn't figure out they were in a TV program. But how does Alda Hertzog come to have internet access while she's locked up? I can't say she came across as quite as threatening as she was probably supposed to though.
“So maybe it's not as sexy as dead poultry in Africa...” l-)

Characters

Name	Actor	Description	Relationships
Mark Benford	Joseph Fiennes	FBI agent	Husband of Olivia Benford Father of Charlie Benford Colleague of Demetri Noh
Olivia Benford	Sonya Walger	Doctor	Wife of Mark Benford Mother of Charlie Benford Apparent future lover of Lloyd Simcoe Colleague of Bryce Varley Operated on Dylan Simcoe
Demetri Noh	John Cho	Junior FBI agent, apparently doomed	Colleague of Mark Benford
Bryce Varley	Zachary Knighton	Formerly-suicidal doctor	Colleague of Olivia Benford
Lloyd Simcoe	Jack Davenport	(Self-described) annoying parent	Father of Dylan Simcoe Apparent future lover of Olivia Benford ?Unwilling co-conspirator of Simon
Dylan Simcoe	Ryan Wynott	Injured autistic child	Son of Lloyd Simcoe Saved by Olivia Benford
Charlie Benford	Lennon Wynn	Winsome child	Daughter of Mark and Olivia Benford Saw D. Gibbons in vision and didn't like him
Stanford Wedeck	Courtney B. Vance	Assistant Director FBI LA Field Office	Boss of Mark Sanford and Demetri Noh
Alda Hertzog	Rachel Roberts	Terrorist
D. Gibbons		Obscure hacker/bomber	Telephone interlocutor of suspect 0 Killer of Pigeon local cop
Suspect 0		Awake during the flashforward	Telephone interlocutor of D. Gibbons
Simon		Sinister guy on phone	Tells Lloyd Simcoe they're responsible for the “greatest disaster in human history” (presumably the flashforward).

Photo catchup

Sun, 18 Oct 2009 20:31:45 GMT

The window display in Forbidden Planet is (or was, if they've since changed it) quite striking.

I can't remember why I took this; I think we thought the no-pregnant-women sign was rather patronizing.

Pattern here. The keys were very fiddly!

Timing bug somewhere-or-other

Sat, 17 Oct 2009 12:58:57 GMT

richard@araminta:~$ cat t.c
#include <stdio.h>
#include <time.h>
#include <sys/time.h>
#include <assert.h>
int main(void) {
  for(;;) {
    time_t t;
    struct timeval tv;

    assert(gettimeofday(&tv, NULL) == 0);
    assert(time(&t) != (time_t)-1);
    if(t < tv.tv_sec)
      return printf("%ld %ld.%06ld\n", t, tv.tv_sec, tv.tv_usec);
  }
}

richard@araminta:~$ gcc -o t t.c
richard@araminta:~$ ./t
1255784018 1255784019.000000

richard@araminta:~$ uname -a
Linux araminta 2.6.26-2-xen-amd64 #1 SMP Thu Aug 20 00:36:34 UTC 2009 x86_64 GNU/Linux
richard@araminta:~$ dpkg -l libc6\*|grep ^i
ii  libc6                                   2.7-18                     GNU C Library: Shared libraries
ii  libc6-dbg                               2.7-18                     GNU C Library: Libraries with debugging symb
ii  libc6-dev                               2.7-18                     GNU C Library: Development Libraries and Hea
ii  libc6-dev-i386                          2.7-18                     GNU C Library: 32bit development libraries f
ii  libc6-i386                              2.7-18                     GNU C Library: 32bit shared libraries for AM
ii  libc6-prof                              2.7-18                     GNU C Library: Profiling Libraries
richard@araminta:~$ cat /proc/cpuinfo 
processor	: 0
vendor_id	: GenuineIntel
cpu family	: 6
model		: 15
model name	: Intel(R) Core(TM)2 Quad CPU    Q6600  @ 2.40GHz
stepping	: 11
cpu MHz		: 2397.654
cache size	: 4096 KB
physical id	: 0
siblings	: 1
core id		: 0
cpu cores	: 1
apicid		: 0
initial apicid	: 0
fpu		: yes
fpu_exception	: yes
cpuid level	: 10
wp		: yes
flags		: fpu de tsc msr pae cx8 apic sep mtrr cmov pat clflush acpi mmx fxsr sse sse2 ss ht syscall nx lm constant_tsc rep_good pni ssse3 cx16 lahf_lm
bogomips	: 4801.28
clflush size	: 64
cache_alignment	: 64
address sizes	: 36 bits physical, 48 bits virtual
power management:
<etc etc etc>

Docking

Thu, 15 Oct 2009 09:18:20 GMT

View Poll: #1471524

Soekris so far

Sat, 10 Oct 2009 16:57:25 GMT

Steve gave me a Soekris net4501. I've been gradually getting this into shape for use as the house router. This page covers the details in case anyone else attempts something similar; some aspects were not entirely trivial.

Testing reveals that it can route data at (at least) 24Mbit/second, which is comfortably faster than the nominal speed of our Internet connection. It's much slower than the nominal speed of the house wireless but actually marginally faster than the measured speed, so I'm not worried about that either.

(“At least” 24Mbit/s because that could be a limit somewhere other than the Soekris.)

Usefully for the testing it turns out that crossover cables are no longer necessary, at least when connecting the Soekris to a modern Mac (and for all I know any other modern hardware).

Make a prediction

Wed, 30 Sep 2009 11:52:23 GMT

View Poll: #1464460

Is it a bird...?

Mon, 21 Sep 2009 20:08:22 GMT

Just walking across the car park at Wimpole when I heard a noise, looked up and said “Woah!”. As chance would have it I had my camera and the right lens.

(That's a Vulcan for anyone who manages not to recognize it.) Later on there was what I'm pretty sure is not only a Hawker Hurricane but in fact the same one as in the picture on Wikipedia.

I had no idea what this was at the time but a bit of Googling reveals it to be a Tiger Moth.

The red arrows were around too and earlier there was some kind of small fast jet doing acrobatics in the general direction of Cambridge. It turns out that the event was the centenary of Marshalls.

Stacks and better translucent windows

Sat, 19 Sep 2009 13:54:42 GMT

I recently upgraded to Snow Leopard.

The first thing I noticed was that it was significantly faster to start up and log in than its predecessor. Unlike certain other operating systems it does seem to be genuinely starting faster too, rather than just getting you to a desktop quickly and then running like a drain while it finishes actually starting up. The Startup Disk system preferences pane is something else that's visibly faster.

Exposé is more tidily organized. In Leopard when the windows shrank to fit they would be rather haphazardly jumbled up. In SL they are in one or more lines, with iconized windows below. Having the iconized windows appear from nowhere is initially a little disconcering (the rest shrink smoothly into place).

Stacks are improved too to the point that I've stopped having my frequently used applications in the dock and instead put just /Applications there and started programs from there. While doing this I noticed that Apple have made a better translucent window, albeit not in a way that they use everywhere. Ordinarily I find translucent windows completely pointless as if the background and foreground are both text-heavy (which they often are for me) the background text distracts from the foreground:

But in a stack, the background is heavily blurred, making it much less intrusive:

Laptop trackball

Thu, 17 Sep 2009 13:24:29 GMT

Can anyone recommend for/against trackballs suitable for use with a laptop?

I'm looking for something that's:

fairly small
won't lose the ball if slung in a laptop bag
ideally, uses bluetooth (though usb might be ok)
ideally, that could be clipped to the side of the laptop

Blue screen

Sun, 13 Sep 2009 13:13:36 GMT

I was trying to get some (any) kind of network filesystem access from my Mac to a Unix box. I didn't have much luck (which is another rant entirely) but I was amused by the “it's ~~broken~~ a PC” icon...

What's the point of ITV?

Sun, 13 Sep 2009 09:44:19 GMT

If the worst should happen, and ITV were dismantled or taken over by an overseas company with less of an obligation to create British programmes, it would leave a huge hole. That really is thinking the unthinkable.

Would it? Frankly I'm not sure I'd notice. The last thing I watched on ITV was Law And Order UK which they stopped showing half way through the series without any announcement as to when or whether the rest will appear. (You might think it would be a better fit on C5 anyway given that's where the US versions show.) I can't remember what would have been the previous thing, and I mostly watch C5 and (to a lesser extent) the BBC.

It would of course help if they didn't, apparently uniquely among TV stations, try to stop you finding out what they were showing.

Last night's moon

Fri, 04 Sep 2009 20:21:44 GMT

New lenses

Fri, 04 Sep 2009 16:44:11 GMT

Canon's new 15-85mm f/3.5-5.6 lens looks interesting to me, and I'll be keen to see a proper review.

I've been using the 17-85mm f/4-5.6 for some time; its overwhelming plus point is the wide focal length range (compared to say the 18-55mm kit lens) and although I've been generally pleased with the results image quality starts to fall when it's wide open (especially when this means both focal length and aperture).

The first of the apparent advantages of the lens is 2mm more on the focal length. This amounts to about a 10% wider field of view (66° versus 73°) and since I keep running up against the 17mm stop, I'm sure I'd use the extra.

The second is another third of a stop in the diaphragm. Again doesn't sound like much but could just be enough on a dim day, and if it pushes the point where things get soft out a bit further that would help too.

It looks like it's physically shorter and fatter, but heavier, than the 17-85mm. The former is certainly good, as with a camera on the back and polarizer on the front the old lens is a tight fit in my camera bag. Fatter means that I'd need a new polarizer.

The obvious downside is the cost, even given that I expect I'll be able to sell my old lens on.

(Yes, I know perfectly well that if you really want high quality, fixed focal length is the way to go, but I have very limited tolerance for (1) carrying around lenses I'm only going to use occasionally and (2) constantly swapping between different lenses.)

In other news they've significantly upgraded their 100mm macro lens with IS and (given the L designation) presumably superior glass, not that I've detected any flaw in the existing one, which I've got a lot out of. I think this will be easier to resist but both of these lenses being announced at the same time has left me feeling like Canon have an eye on my wallet in particular l-)

50 things that are being killed by the internet

Fri, 04 Sep 2009 11:50:28 GMT

50 things that are being killed by the internet (or more accurately, by modern communications technology). Which do you miss the most? The least? What have they missed?

Bin there done that

Thu, 27 Aug 2009 22:12:46 GMT

BBQ Photos

Thu, 27 Aug 2009 20:17:16 GMT

Small PCs

Mon, 24 Aug 2009 16:42:16 GMT

I'm planning to get a Soekris net5501 to the be the house router (so: IP routing, NAT, DHCP client/server, secnet VPN, DNS). Before I do, are there any similar devices I should consider as possible alternatives?

What I need is:

Silent or at least very quiet
At least three ethernet ports (house LAN, house wireless, upstream)
Must be able to run Linux without undue hassle

Nice to have but not necessarily essential:

Serial console (but I can faff around with a keyboard and monitor if it comes to it)
Compact flash storage (I have a spare 2GB card, which is unrealistically small for my camera)
x86 architecture (my plan is to transplant an existing lenny install onto it)