Longmead River

There was a gang out clearing out the stream - a tributary of the Hogsmill - which runs down Longmead Road this morning. A gang from a van which I thought said Paul Draper, river and fishery management.

Point of interest one. Thames Water, which I assume is the responsible authority, has contracted out some of the periodic clearing out of smaller water courses. With a contract which seems to be a little vague about whether the rubbish so cleared out should be shoved in the hedge or taken away.

Did it include water quality - given that I noticed a white plastic boom across the stream the other day, the sort of thing they use for trapping floating or foaming pollutants?

Point of interest two. Neither Bing nor Google turned up any likely looking contractor of this name and while LinkedIn has ten or twenty Paul Drapers on their books, none of them look any more likely. Unusual these days for even quite modest contractors not to be web visible. I don't think that I got the second name wrong as I associated to the people who make hand and other tools. See reference 1.

Point of interest three. The large trolley which I have been eyeing up in the stream for a month or so now, chickening out to date on the grounds that there was no convenient tree to tie a safety tether to, will probably not be there tomorrow. See reference 2. But I shall look out for it: shoved in the hedge?

Reference 1: https://www.drapertools.com/.

Reference 2: http://psmv3.blogspot.co.uk/2017/11/a-challenge.html.

Wilde two

Following our excursion to Wilde a couple of months ago, noticed at reference 1, back to the Vaudeville for some more last week, in the form of the rather more popular 'Lady Windemere's Fan'.

On this occasion we took our picnic in the Clore Ballroom, rather than in a flanking terrace, mainly surrounded by mothers with their toddlers. Some of whom were energetically scooting about the place: probably forbidden but harmless enough in the circumstances. Mothers who made a fair amount of mess with their picnics, but who tidied up after themselves.

On for pre-theatre refreshment in the Nell Gwynn in the splendidly named Bull Inn Court. A proper old fashioned boozer. A bit cramped, lots of brown wood, a bit scruffy and very little in the way of food. Furthermore, the barmaid had been a maid in the last excursion, 'A woman of no importance'. I was tactless enough to explain that we thought the play a bit long. She explained that she had been one of the maids, which I had forgotten about until BH reminded me, after the event, of the rather jolly musical numbers between some of the scenes. Also that she was paid £650 or so a week for the duration which she thought was quite good money, so presumably better than barmaiding.

Later on I tried to remember what 'A woman of no importance' was about, and completely failed. I tried a refresh with the long plot synopsis offered by Wikipedia, but I did not have the patience to wade through that. So all that I am left with is that it was another play involving a mix-up over parentage - as was, as it turned out, the one to come. A subject which was perhaps all the rage at the time Wilde was writing - or perhaps close to his mixed up heart. Perhaps it was common for bored housemaids, when they were not having fantasies about white slavers, to have fantasies about being the natural child of some prince or other.

Another clever production, nicely staged and dressed. Better than the last one, but still a little too long. And the third scene, the boozy club men scene, seemed rather forced somehow, rather in the way that the gangster failed in the first half of the birthday party. See reference 3. More clever song and dance in some of the intervals between scenes, one including a star performance by a maid. The sort of thing which might, I suppose, have featured in the music halls of Wilde's day.

After this second outing, our verdict was that Wilde had worn quite well, but was past his sell by date. No need to see any more any time soon.

Out to get a little lost in the area once known as the Adelphi (I think from the Greek, after the three Adam brothers), a losing noticed at reference 5, and then onto the Villiers Street branch of L'ulivo, a place we have used occasionally in the past, just about seven years ago according to reference 6. Quite possibly a family-run independent, with just the two branches.

Presentation, service and ambience good. Garlic bread a little salty, otherwise good. Spaghetti Bolognese adequate, rather than good. Rather too wet for my taste and not enough spaghetti. Perhaps it was a mistake to take one of the cheapest main courses on the menu. I was also reminded of the BH story about Italians thinking it odd how much pasta English people like to take with their sauce. Supplementary bread poor, fresh out of a machine bakery. Tiramisu good, better than I have had for a while. Appeared to have been made in a shallow bowl, with half a bowl being a portion, rather than the more usual slabs of commercial confectioners. A new us white wine from the 2015 vintage called Valentino which we rather liked. It even had a real cork, which caused the young waitress a bit of bother; clearly no longer used to such things. And, contrary to what Wikipedia might say, the bottle said something about Tuscany.

Back across Hungerford Bridge to Waterloo, to be offered a seat by a smart young lady in the rush hour train to Epsom. It took me a little while to realise what she was on about, but I accepted in the end. Not too proud to sit on this occasion!

PS 1: the illustration found by Bing, possibly from the first production of the play, back in 1892. Possibly the St. James's Theatre which is now Webber's Other Palace.

PS 2: looking at the tickets again this morning, I was reminded of the irritating practise whereby the price you pay for your tickets is significantly different from, invariably more than, what is called the face value. I suppose that theatre companies have found, like insurance companies, that flexible prices is more profitable than fixed prices. Irritating though the former are to people like me. A practise which, I am pleased to say, does not seem to have percolated through to concert halls.

Reference 1: http://psmv3.blogspot.co.uk/2017/11/wilde-one.html.

Reference 2: http://www.thenellgwynne.com/.

Reference 3: http://psmv3.blogspot.co.uk/2018/01/the-birthday-party.html.

Reference 4: http://www.lulivo.co.uk/.

Reference 5: http://psmv3.blogspot.co.uk/2018/01/fake-23.html.

Reference 6: http://pumpkinstrokemarrow.blogspot.co.uk/search?q=l'ulivo.

Trolley 121

This trolley from Sainsbury's being captured outside Epsom Station, at the junction of Station Approach with Waterloo Road, more or less opposite the Co-op. It must have been a fair way to walk the thing, loaded, from Kiln Lane.

Not sure whether I would have taken it back had there been a deployed wheel lock, requiring the trolley to be tipped up. There is a passage which by-passes the High Street, but it is still a fair way to go tipped up, earlier remarks on how good I was getting at it notwithstanding.

PS: regarding the University of Creation stash from which the trolleys 120 came from, I passed three young ladies yesterday, happily walking their lightly loaded trolley down East Street, away from Kiln Lane. Almost certainly creationists and it looked most unlikely to me that they would even think of returning it. A pity that I was not walking a trolley myself on that occasion, which would have made an opening for a bit of banter.

More oxtail still

For the record, we have just consumed the third oxtail of the winter season, with the second having been noticed at reference 1.

Usual drill, that is to say two coarsely chopped onions in bottom of pyrex dish, then three pounds of oxtail on top of that, quite possibly making up one entire oxtail. Cover and cook at 100C for six and a half hours. Turn. Further 30 minutes. Drain. Further 30, with oven off for the last 15.

Meanwhile, prepare and cook boiled white rice and boiled crinkly cabbage.

Very good, just the right amount of chewy succulence. Better than the six and three quarter hours version last time. Four small pieces left to nibble over Midsomer Murders later. Bit of veg. left with which to top up the left over chicken soup from yesterday. With lentils as well as chicken, naturally.

Rounded out with dried apricots, gently stewed.

Reference 1: http://psmv3.blogspot.co.uk/2017/11/more-oxtail.html.

News items

First the House of Commons. The Commons is finally getting around to thinking about what to do about their house, in imminent danger of falling down, catching fire or worse. But I think that the chances of their opting for a modern legislative chamber, along the lines of more or less everywhere else, is about zero. Still locked in our glorious past and quite failing to understand that the present is not very glorious at all.

Think what an opportunity they missed on that plot, recently vacant, next to the new US embassy at Vauxhall. A new house there could easily have been up and running by now. And the old one could have been sold on to Merlin (reference 1) as an entertainment.

Second the House of Cladding. It seems that there is a block of flats up north, clad in the same sort of cladding as Grenfell Tower.  However, this tower was a private speculation which has been sold on, on leaseholds. The sort of leasehold whereby the leaseholders are responsible for maintenance, internal and external. Rather in the way that the leaseholders of the low rise blocks of flats on my own estate are jointly responsible for the considerable expense of retiling their roofs from time to time. So are these particular leaseholders entitled to any sort of help from the public purse?

It is not clear to me that they are. This was a risk they took on when they bought the leasehold. Maybe they could get some traction from negligence by the speculator - although that must be weakened if the speculator could sustain the claim that this fire risk had not been invented at the time of construction and so was not down to them.

Third the Dorchester. How much lucrative charity trade have they lost by being caught out hosting sex parties for the great and the good?

In which connection I have often wondered about the morals (or perhaps the virtues) of having fancy balls and banquets at such places, while flying under the charity banner. The amount of money passed on to charities was often modest when compared with their costs.

Reference 1: https://merlinentertainments.biz/.

Footnote to perfection

Computers really will soon rule the world.

This morning's bit of evidence being the products from Neuratron (reference 1) called PhotoScore and NotateMe 8.

Using a demonstration version of the product - found, loaded, up and running in seconds - I was able to scan in a short bit of music from 'The art of music' noticed yesterday (reference 3) and play it on a virtual piano. As far as I could see its transcription from the not very good image I had given it was note perfect. It had turned an image into scored music which I could read (and edit) and it could play.

No doubt, given time, I could do all kinds of wizardry. But I shall have to pause to think whether there is not a lower tech solution to my need to play the musical examples from the book. Which, to be convenient under this first option, would really need a second copy of the book which I could take to pieces for ease of scanning, while keeping the first copy in my hand.

Option 2, activate the recorder in the roof. An instrument I could play a little, sixty years ago.

Option 3, visit Maplins.

Option 4, download one of the many piano keyboard emulators which seem to be out there on the internet, some to be paid for with money, some very fancy (reference 2) and some to be paid for with advertisements.

We shall see.

PS: pleased to see that Neuratron is a British company, based in not so far away Blackheath. We can still make stuff. Not dead yet.

Reference 1: http://www.neuratron.com/.

Reference 2: https://www.pianoteq.com/.

Reference 3: http://psmv3.blogspot.co.uk/2018/01/perfection.html.

Perfection

Quite by chance, I came across the book at reference 1 at the second hand book stall at the bottom of the stairs to the museum at Bourne Hall, a stall from which I pick things up from time to time.

Being the eightieth volume in the 'International Scientific Series' published by Kegan Paul, Trench, Trübner and Co., Ltd. I learn that Kegan Paul was name of one rather than two persons, a person who went to Eton and then to Exeter College, Oxford and that Hubert Parry also went to Eton, took degrees from Cambridge and Dublin, was a fellow of the same Exeter College and was made a baronet for his services to music. Some things just don't change.

An interesting and varied list, and one supposes that the books were the sort of thing that the part educated bought, partly for interest, partly to better themselves. From the golden age of reading rooms, mechanics institutes and libraries. See reference 4 for one of our local examples.

And as it has turned out, this book might be just the thing to sort out my lack of knowledge of the theory of music, something that I poke at from time to time, without much lasting effect. To the point that, having for perhaps the third or fourth time worked out how a major scale worked, I decided it was time to make my own diagram, included above. Perhaps it will now stick. Note the way that the middle notes of an octave arise in a more or less natural way, with the ear recognising and liking the simple intervals involved, while the outer notes are more contrived.

All helped along the way with reference 2 and various handy summaries for things like 'middle C' from Wikipedia. So far the book seems remarkably up to date, considering that it was written well over a hundred years ago and so I share a few other snippets.

Back in the beginning, everyone had octaves but the original big divide was between five note scales and seven note scales, seemingly associated with whether there was a preference for downward fourths rather than upward fifths. With a lot of scales, even now, going down by convention, whereas ours go up.

The Greeks, the Europeans, the Arabs, the Persians, the Indians, the Javanese, the Siamese, the Chinese and the Japanese were all into scales of one sort or another. All peoples who were more into melody than rhythm, with the savages of the south being more into rhythm than melody. This according to Parry, back in 1900 or so, with the invention of the savage mind by Lévi Strauss not coming until more than half a century later.

The Europeans moved on from just temperament to equal temperament, got very good at harmony and counterpoint and produced the Mount Everests of the musical world. But before that, there were various papal edits (aka decretals), particularly from popes called Gregory, about how things should be done. Hence Gregorian chants. It seems that at one time, authorities around the world took a keen interest in these matters.

Reference 1: The evolution of the art of music - Sir. C. Hubert H. Parry – 1905.

Reference 2:  Chamber’s Encyclopedia – various authors - 1959. The article on temperament.

Reference 3: La Pensée sauvage - Claude Lévi-Strauss - 1962.

Reference 4: http://www.leatherheadca.org.uk/.

Second attempt

First batch, first side, two rings removed for turning

Just about a month after the first attempt, a second attempt at crumpets yesterday, using the same recipe but getting the yeast right (at a tablespoon rather than a teaspoon) and frying in rapeseed oil rather than in butter.

The yeast, sugar and milk was left for about 15 minutes, and the batter for about 45. After which the batter was a most interesting texture, wet and frothy.

Ladled into the rings in the heated pan (hob on 4) to a depth of about three quarters or about half an inch. Some rising, but not so much as to stop the rings coming off neatly enough. Holes after the way of shop crumpets.

In the event, a bit too brown on the outside and a bit too soft on the inside. But entirely eatable.

Second batch, second side

I should by this time have known better, but more or less filled the rings up for the second batch, which meant that they overflowed. And with the hob down to 3, still too brown on the outside and too soft on the inside.

Third batch, second side

Not much batter left for the third batch, so thinner perforce. After the event, I read that Delia Smith suggests using a tablespoon to the crumpet - a lot less than I used - while Mary Berry suggests 2cm - a bit more. While Delia Smith also says that I should get a dozen crumpets - which does not add up - but then I have often wondered about the standard of checking and proof reading in these fat books knocked out by celebrities. Or at least knocked out in their names.

All quite eatable and all eaten hot with butter, but I think next time I will try 1cm thick and turning the hob down to 2 - but giving it a good time to come to temperature. And I need to work on my turning technique. And BH suggests cutting the milk with water, the fact that the milk is already skinny green top notwithstanding. 

In taste, more like English pancakes, the sort you take with lemon juice and castor sugar, than Shop crumpets, although slower cooking may change that. In any event, I am sure things will improve with practise, just like they do with pancakes.

PS: what would have happened to the batter, the ingredients for which were not that different from those for bread, had I tried baking it? Perhaps something else to try, a something which ought to result in a lot less washing up.

Reference 1: http://psmv3.blogspot.co.uk/2017/12/crumpets.html.

Haig one

I recently acquired, for the rather large sum of £20, a copy of Duff Cooper's two volume biography of Haig, a book of which I had been aware for a while but had never got around to.

From a sale at Epsom Library, a nicely printed hardback from Faber, which must have been much read over the years as it has been rebound in library blue by Riley, Dunn & Wilson of reference 1. A pretty good job except that some of the pages are now slightly crooked and some of the sewing is more visible than it ought to be. However, read just three times in the last 15 years or so, so perhaps Surrey Libraries did not time their rebinding very well.

As it happens, my mother had a copy of Duff Cooper's wife's three volumes of memoirs, which I then had for many years, unread, eventually culled. I am fairly sure that I also once owned Duff Cooper's own memoirs 'Old Men Forget', also culled, quite possibly unread. And then I once owned their son's (John Julius Norwich) three volume history of Byzantium, both read and culled.

This post to notice my completion of the first volume. An easy read, if it was also easy to get lost in the maze of names of places and units. Much better maps would have been helpful.

Born in 1861, the youngest of the many offspring of the inventor of Haig whisky. I have not yet found any mention of his contemporary Mrs. Greville, the heiress of the inventor of McEwan's Export, another Scot who made it to the top of the social scene down south, helped along by money from trade. Known to us as the castellan of Polesden Lacey and as a very good friend to Osbert Sitwell, both much noticed in previous posts.

Not thought to be clever but nevertheless passed through Oxford and Sandhurst to join the cavalry, spending a lot of his time there being very good at polo. But he must have been able in other ways, because serving in India, the Sudan and the South African war he rose rapidly, making general officer rank in time for the first world war and becoming commander in chief of the British expeditionary forces at the end of 1915.

I finished the book with a much better opinion of the first world war generals than I started. They did what they could with the men and weapons available. Avoiding, stopping or otherwise curtailing the dreadful slaughter was a job for the politicians not the generals.

The slaughter of the Battle of the Somme being one of the charges against him. In his defence, Duff Cooper points out first that the battle was fought to take the pressure off the French at Verdun, who might otherwise have collapsed - and while a peace on German terms in 1916 might have been better than what happened 30 years later, it might not have been that great. Second, that the time and place was of French choosing. Third, that while the slaughter was dreadful, what we bought for that slaughter was the destruction of the German army, which might have fought on for another two years, but was never the same again. An opinion which I believe has become fashionable once again.

I associated to remarks in Churchill's book about Marlborough, to the effect that Marlborough believed that to win, you had to smash the main force of the enemy, probably by frontal attack. Costly, but effective, certainly in the run up to Blenheim. In the present case the main force of the enemy was in France and Belgium and footling about in the eastern Europe or the Near East was beside the point and a waste of scarce resources.

I was reminded that, certainly up to the end of 1916, the war was much more a French affair than a British affair. They had a far bigger army in the field, it was their country and they called the shots. Leaving Haig to play a tricky game between the French, his army boss back in London, his political bosses back in London and what he thought should actually be done. Duff Cooper passes on a few harsh words from Haig about politicians who talk well in committee but who know nothing and who, all too often, got their mistaken way.

There was an interesting illustration of the limits of his power. Some time before the Somme, Haig had heard of a new infantry tactic which involved crossing open ground in short bursts, rather than in long lines spread out across a long front. He tried to persuade his generals to give it a go, but failed and did not feel able to press the matter. In the event, the new infantry tactic became the norm rather later on, after many long lines had been mown down.

I was struck by the number of generals who were killed, certainly up to the end of 1916. This in contradiction to a claim that I had once heard that no-one above the rank of major had been killed. All the senior officers were safe, sipping good claret in their châteaux behind the lines. On at least one occasion, careless copy from a journalist resulted in such a château being shelled, with many of its occupants being killed. Killing their generals was always a good way to disrupt the enemy and both sides did as much of it as they could manage. None of this Wellington nonsense at Waterloo of it being none of his business to take pot shots at Napoleon.

But I do remember that I tried to check this claim, which I did not believe, and completely failed to turn up any statistics about casualties by rank. Perhaps I would do better if I tried now.

I was also struck by how old many of the generals were, with a lot of them being well over sixty, perhaps inevitable given the long period of relative peace which had preceded the first world war. But far too old to be able to cope with the work load and responsibility of a general in war time. In which connection I was interested to read that one of the ways that Haig coped was by attending Divine Service (Scottish variety) and taking his prayers seriously. Which I might not approve of, but what other crutch had he got to lean on, in the circumstances of the time? Neither drink nor drugs would have been very good alternative plans.

Finally, following the read noticed at reference 2, it was interesting to read here of the Indian Army furnishing one of the half dozen or so army corps in the field at the end of 1914. With an army corps at the time being two or three divisions, perhaps 50,000 men in all. There was also a splendid anecdote about what I assume was a Sikh general, another cavalry man, asking his chum Dougie when they going to able to have a proper charge. I believe all the combatants on the western front kept some cavalry ready and waiting for that opportunity for the duration. While those on the eastern front actually got to use theirs.

PS: later: I think the unread qualifier to 'Old Men Forget' is probably correct. Diligent search of the blog reveals just the irrelevant reference 4.

Reference 1: http://www.rdw.co.uk/.

Reference 2: http://psmv3.blogspot.co.uk/2018/01/partition.html.

Reference 3: 'Old men forget'. King Henry V, Act IV, Scene III.

Reference 4: http://pumpkinstrokemarrow.blogspot.co.uk/search?q=Diana+Duff-Cooper.

Stuffing

No stuffing Christmas past as we failed to eat fowl and the first notice I can find is more than two years ago, at reference 1. Surely we have stuffed since then?

Leaving history aside, we did have fowl yesterday and I thought, rather late in the day, to take it with external stuffing, the sort made in the sort of white enamel pie dish which was common when I was a child and is becoming common again, albeit in a smaller size, in the sort of middle range restaurants which we use. See, for example, reference 5.

Too late in the day to arrange for some white bread crumbs, so as at reference 1, reduced to using brown, of my own baking. Along the way finding that an ordinary fork is just the thing for stripping the maximum amount of crumb out of the shell of crust.

A modest amount of celery, maybe a cup when chopped.

Two medium onions, maybe two cups when chopped.

A couple of tablespoons of chopped hazelnuts, probably bought entire from the Epsom branch of the people at reference 4, good at that sort of thing.

A tablespoon of freshly chopped sage, a tablespoon of dried - our rather tired sage plant not being up to both spoons.

Half a dozen black peppercorns, freshly pounded. Pound until they can be smelt at six inches.

Two eggs.

Fill tin, cover with wet bacon from Sainsbury's. Bake for around an hour and serve with fowl.

Quite good, gone within four hours of serving, eating rather well cold. But I still think I would have preferred the white version.

PS 1: I felt sure that I had mentioned the people at references 2 and 4 before, but it took me a while to track them down. Searching the blog archive with 'organico' turns up nothing while 'epsomorganico' produces reference 2. Maybe one day I will get it clear in my head about which searches do things like startswith and contains by default, and which don't. And I think that 'organico' is the name of a product rather than the name of a shop.

PS 2: the business with startswith and contains is probably a good example of the requirements business noticed towards the end of reference 6. Microsoft, Google and other suchlike people have probably worked hard to tune the default search settings for the various contexts in which they offer a search feature. Which is good up to a point, but irritating when doing something a bit tricky and one gets confused about which search does it which way. An irritation which would not arise if they all did it the same way. Perhaps all offering not too tricky ways around the defaults, perhaps offering options like 'startswith=yes' and 'endswith=no' which can be tagged onto the end of the search term, just to make sure.

Reference 1: http://psmv2.blogspot.co.uk/2015/05/stuffing-external.html.

Reference 2: http://psmv2.blogspot.co.uk/2015/07/stuffing.html.

Reference 3: http://www.organico.co.uk/.

Reference 4: https://www.facebook.com/epsomorganico.

Reference 5: http://psmv3.blogspot.co.uk/2017/01/sewing.html.

Reference 6: http://psmv3.blogspot.co.uk/2018/01/on-similarity.html.

Computers that play chess

Figure 1

From the newsletter published by the people at reference 1, I learn about the work reported at reference 2. Getting computers to beat grandmasters at chess is clearly not good enough any more.

People have been trying to get computers to play board games such as Checkers (also known as Draughts), Chess, Shogi (a more complicated variation of chess) and Go for a very long time. See, for example, the much cited paper from 1959 about checkers at reference 4.

All these games are two player board games played on a rectangular grid, from eight by eight for Chess to nineteen by nineteen for Go. One sort of piece in Checkers and Go, a dozen or so in Chess and Shogi. Perhaps black pieces for player one, white pieces for player two. A repertoire of permitted moves, which might involve the capture of one of the opponent’s pieces. Games can result in a win, a draw or a loss. At higher levels, played with rules about time. All the games are open, with both players having perfect knowledge of the rules and of the play so far, memory permitting. Lots of history. Lots of competition.

All these games are discrete and finite. There is nothing like, for example, position in a war game, where a unit may move to any position in game world, not constrained to some small number of positions on a grid, maybe not even to positions on some two-dimensional surface. Where the attributes of units are orders of magnitude more complicated than those of the pieces in these games. And moves alternate in an orderly way – which may not be the case in a war game, played along realistic lines. Three properties – discrete, finite and alternate – which make them attractive targets for computer programs. But with the catch that, with the possible exception of checkers, there are still far too many possibilities for brute force, for the exhaustive search of possibilities, using any computer likely to be available any time soon. Something more cunning than brute force is needed. And even if this were not true, cunning is more interesting, more entertaining than brute force.

In the event, most of the many computer programs used to play these games have been organised in a straightforward way, at least at the highest level, with an evaluation function and a search. The evaluation function says whether this or that position is good or bad. The search function looks ahead from the current position and uses the evaluation functions to score all the possibilities and so to decide what to do.

Evaluation functions are often defined by means of features and weights. One has a long list of features that any given position may or may not have and one has a long list of weights. And one has some function which combines the features and the weights to give a score. Frequently the function is simply the weighted sum, the scalar product. And one of the tasks of the program is to learn the weights that work, that result in wins.

Given the convention that player one on his move wants to maximise the evaluation function and that player two on his move wants to minimise his (usually, in these programs, the same function), the search is usually some variation on the alphabeta variety of the minimax search. Wikipedia knows all about these matters and one might also take a look at reference 6.

Huge amounts of effort have been put into building and tuning these two functions, using huge amounts of domain specific knowledge. Some of the programs include libraries of beginnings and ending and are able to concentrate on what lies in between. Some of them include cunningly crafted algorithms to deal with particular problems and issues. Some of them try to encode the wisdom of the masters in some other way.

Huge amounts of time were put into training these programs, training which often made use of the libraries of expert chess games which are now available. Other programs trained by playing themselves or each other.

All this reached the point, relatively recently, where computer programs supported by large computers could beat world ranking humans.

Alphazero

But now, a lot of this has been swept away, by the success of the people at Google with their Alphazero program, the program reported on at references 1 and 2.

All this program knows about is this class of games in general and the bare rules of the games that it can play in particular. The only slightly odd looking bits of information that it needs are the length of a reasonable game and the maximum length that it need bother with. It has no libraries of grand master games or anything else. It has no cunningly crafted algorithms doing special, game specific things. And it can learn to play world standard chess, or whatever, more or less overnight – with Figure 1 above illustrating the convincing results of testing this allegation by competition with other world ranking programs - including here the previous incarnation of itself.

One of the interesting things that this program does is generate winning moves that a human player would regard as bizarre. In which it seems to be going beyond what humans have managed so far.

I did not notice anything about how a big a computer is needed to run this program, how it compares, for example, with the sort of thing needed to forecast the weather or make sense of the goings on in the large hadron collider.

Bits and pieces

Along the way, a lot of work has been done on something called TD(λ) learning, where ‘TD’ is temporal difference and ‘λ’ is a parameter which controls the use of information from past times. With the issue being how relate outcomes, good or bad, to actions (in this case moves) which might have taken place some time ago and with TD(λ) learning being all about learning as you go along, without waiting for the final outcome. A tool which has been around for a long time, successful a long time ago on backgammon, but now seen to be of wider applicability. See reference 3.

Along my way, in writing this post, I came across the work of John Holland, in particular his bucket brigade algorithm and his Echo program. Which last now seems to live at the interesting outfit called the Santa Fe institute of reference 7. But both Google and Bing were oddly ineffective in turning up a proper description of the bucket brigade. So far, the best they can do is the introductory reference 8, as it happens the work of a fellow blogger.

Feature vectors also feature in the language work noticed at reference 9 and the image work noticed at reference 10 – both involving, as it happens, engineers from Google.

Following reference 5, I was interested to see layers coming into play with Alphazero, with information being stored on a large number of planes, with each plane carrying one bit, or perhaps one byte, of information about each position on the board.

I do not pretend to understand much of what the Google team have done. But I believe that much of their progress has stemmed from their having learned how to train neural networks with lots of layers, with previous efforts having been stuck at a very small number of layers. With more layers neural networks can do much more complicated stuff, stuff that one could probably not contemplate using the hand crafted code of old. I imagine that this includes better training for evaluation functions.

I also believe that they have moved decisively away from the alphabeta searches which have dominated the field hitherto, towards a more statistical approach. There is talk of methods involving the phrase ‘Monte Carlo’.

I associate to the march of progress in other spheres, with years of work, satisfying careers even, being swept away by some invention or other. With the way of life of the likes of Silas Marner being swept away, in his case, by the advent of the mechanical, the industrial loom. And of Edward Casaubon being pipped at the post by some Germans from some place called Tübingen.

Conclusions

Computer programs can now beat humans at these kinds of games and to that extent the glamour has gone out of trying to write them. But the Google team will no doubt be looking to leverage their inventions in other fields of endeavour.

While I wonder what lessons can be extracted from the way in which Alphazero plays its games. What does Alphazero tell us about these games and the way in which they should be played? Or if it is still too much down to the brute force of big computers to be of much interest to a human player, what does it tell us about how a brain might do other things?

It is not enough, at least not enough for me, to say that we have built a black box which solves the puzzle. I want the know all about the workings of the black box. Analogies might be the structures postulated by Freud – such as the id and the superego – to explain the workings of the mind, the chemical structures which do explain the workings of chemicals, without needing to worry all the time about the myriad activities of sub-atomic particles, or the diagrams which are drawn to make sense of what one sees down a microscope.

PS: Google’s personnel people clearly search far and wide for their engineers, if the names on the paper referenced below are anything to go by.

References

Reference 1: http://www.kurzweilai.net/.

Reference 2: Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm - David Silver, Thomas Hubert, Julian Schrittwieser, Ioannis Antonoglou, Matthew Lai, Arthur Guez, Marc Lanctot, Laurent Sifre, Dharshan Kumaran, Thore Graepel, Timothy Lillicrap, Karen Simonyan, Demis Hassabis – 2017. 

Reference 3: Reinforcement learning in board games - Imran Ghory – 2004. Written as a final year dissertation at the University of Bristol and which, inter alia, contains useful background and tutorial material.

Reference 4: Some studies in machine learning using the game of checkers - Arthur L. Samuel - 1959. But beware: there are a paltry 19 footnotes and references. Clearly not a proper academic at all.

Reference 5: http://psmv3.blogspot.co.uk/2018/01/an-introduction-to-lws-n.html.

Reference 6: http://chessprogramming.wikispaces.com/Alpha-Beta.

Reference 7: https://www.santafe.edu/.

Reference 8: http://vinodwadhawan.blogspot.co.uk/2013/05/79-bucket-brigade-in-john-hollands.html.

Reference 9: http://psmv3.blogspot.co.uk/2017/11/reading-brain.html.

Reference 10: http://psmv3.blogspot.co.uk/2017/11/more-google.html.

Cheese

Earlier in the week off to Borough to replenish cheese supplies from the branch of Neal's Yard Dairy there.

Now it so happens that there have been a number of selfish parents of children at the nearby school (Stamford Green) who, rather than park their tractors a few yards away, or even, heaven forfend, walk their children to school, park their tractors all over the grass verges, doing serious damage at this time of year, when the ground is wet. The headmistress of the school has, to her credit, tackled parents about this in the past, but some of them are still at it. And one was at it on this day and I tried tackling him, not thinking to take a picture of his car first, which would have been a useful fallback. A small scruffy man, possibly Italian, told me to mind my own business and was not impressed when I suggested that as a resident it was my business. And when I asked him how he would like it if I were to key his car or trash his front verge, he became abusive and I did not try to face him down. Chickened out I suppose. I doubt whether I have done any good, but one never knows: people of his sort are unlikely to back down to your face, but maybe he will give the matter some thought later.

Arrived at Epsom to find that trains to Waterloo were in a bit of a state, with the announcer passing the buck by telling us at regular intervals that it was all due to train crews obeying the instructions of the signal crews. I was not much affected as a very late running train then rolled into Epsom, picked up a few passengers and then proceeded, more or less non-stop to Waterloo. At least there were no station stops.

Pulled a Bullingdon from the pole position at the top of the ramp, only to discover that the luggage strap had snapped. But not a problem as tying my bag on with the loose end of the strap worked OK.

Slightly hairy cycling down the ramp in the dark, busy with commuters on foot and on wheels, travelling both directions. Managed not to hit anyone. I did not press button red for damaged cycle at the end of the trip of just over ten minutes.

Cheese bought, I made a picnic of bread and cheese (from the stuff bought from Waitrose in Kingston, noticed at reference 1). Then onto the Barrow Boy and Banker for a spot of their white and some discussion of the different ways in which internal conflicts arise and are resolved in government departments and insurance companies, before heading back to Waterloo, a trip of just under ten minutes.

Slightly hairy on this leg on account of a coach which declined to overtake me down most of the length of Stamford Street, despite there being room, but which then did overtake me at the Waterloo end of the street, stopped at the crossing well clear of the pavement and then opened up his passenger door just as I came up his nearside. Luckily I had enough seconds in which to react.

Position 3 at the top of the ramp.

Ten coachers - trains not buses - now seem to be common on the Epsom line, perhaps the big contribution of the new franchise holder, which means that the doors of the last coach now open right on the stairs down from the platform, a quicker exit to the taxis.

Except that I got into a muddle with my tickets. Luckily the girl on the barriers was a regular, knew me for a regular guy and waved me through.

Reference 1: http://psmv3.blogspot.co.uk/2018/01/kingston-upon-thames.html.

The cello affair

The cello affair continued with Messrs. Queryas and Tharaud at the Wigmore Hall last week, a concert which involved a cello from Turin that was older than any of the music being performed. A cello which was described in the (badly dated) program as being racé, loosely a thoroughbred, as in horse. See reference 1.

Two striking girls on the tube to Green Park, both wearing substantial red coats, which seem to be the fashion this year. The first was a European girl, with a dark red coat, was thin and about six feet tall. The second was a Chinese girl with a bright red coat, open over pale blouse and white trousers. But what was most striking was her face makeup. With most of the face having been lightened, matching bright red lips and a hint of darker blusher about the cheekbones. One wondered how long it had taken to put together - but the result was probably worth it!

Out to find no light show, beyond an illuminated fish cycling through Berkeley Square, advertising the nearby sexy fish restaurant. A place which I do not expect to be gracing; not being young enough, not sexy enough and not rich enough. See reference 4 for details of their private dining experience. Better than the nearby Dorchester?

There was at least one cello in the cloakroom, and I dare say there were more cellists in the auditorium, certainly if the learned conversation which went on behind me was anything to go by.

On the other hand the barista did not know that a ham sandwich (not bad at the Wigmore) counted as a bagel as far as the till was concerned. An error which cost her at least 30 seconds.

Flowers two sorts of dark red, green backing with just a touch of white.

The Bach (BMV 1028), the Schostakovich (Op.40) and the Brahms (Op.38) were all new to me and all very good, although I found the piano a little loud in the first half of the Bach, not leaving enough room for the cello. All very different, but all very much what it said on the tin, that is to say a recognisable part of the oeuvre of the composer concerned. Plenty of echoes. Not so sure about the Berg (Op.5, arranged for cello and piano), but then it was not very long. Sufficiently obscure that an iPod or some such had to be deployed for the cello part - with old-speak paper making do for the rest of the time.

An arrangement of Hungarian Dances 11 and 5 by way of an encore. The second preceded by the first alcoholic toast (champagne) that I have seen on the stage of the Wigmore Hall - a toast to their recently released CD, fortuitously on sale and signable after the concert. All very French and the audience - including me - loved it.

Exit to the balloon noticed at reference 3.

PS: listening to YouTube today, I think I prefer the cello version of No.11 to the more usual two pianist one piano version.

Reference 1: https://tarisio.com/. Perhaps these people, having featured in the program noticed at reference 2, have got themselves a regular slot in the new look programmes.

Reference 2: http://psmv3.blogspot.co.uk/2017/12/sonatas_30.html.

Reference 3: http://psmv3.blogspot.co.uk/2018/01/balloon.html.

Reference 4: https://www.sexyfish.com/.

Reference 5: http://psmv2.blogspot.co.uk/2015/07/the-cello.html. Probably the last time I heard Queryas.

Fake 23

For once, gmaps a little out of date in central London, but this snap is believed to have come from a recently refurbished building at the top of the street which is quaintly named York Buildings, the street which runs southeast from its junction with John Adam Street. This last being the home of the once very grand Adelphi Hotel, used for Poirot films and at one time, possibly no longer, as a headquarters building for the DHSS or perhaps the DSS (now further rebadged as the DWP).

This building is included as a fake for the black steel girder, very Tate Modern turbine hall, worked into the decorative scheme for this front door into something or other. Although given that the steel might quite possibly still be structural, whimsy might be a fairer term than fake.

PS 1: I had thought the Tate steel had come from Kirkcaldy, a relative of the testing establishment of the same name in Southwark Street, but I now find that the latter is actually called Kirkaldy, so perhaps no relative at all. At least I now know that the latter is, or at least was, open on Sundays, so maybe I will get to visit. While the best that the blog can do is reference 2, although I feel sure I have snapped the steel in question at some point. Hopefully I will remember to check when next at the Tate.

PS 2: we had forgotten about the very grand carriage entrance on the river side of the Hotel. At least what might once have been a hotel, but is now office space. See reference 1.

Reference 1: http://adelphibuilding.com/.

Reference 2: http://psmv2.blogspot.co.uk/2015/09/basement-3.html.

On similarity

Introduction

Prompted by the book, still only half read, at reference 1, we have been attempting to build a model in Excel VB which does some of the work of a neural network – without having to go to the bother of actually building one. Or of learning how to use of one the many freebies which we imagine can be turned up by Bing or Google.

This model includes a number of nodes, which may be active or inactive, with degrees of activity between. Nodes have names, names which are often ordinary English words, like cat or dog, are strings in geek-speak and which may not be unique.

This model also includes the idea of similarity between strings: this string is like that string, expressed as a function taking two strings as arguments and yielding a small, non-negative real number by way of result. We might use the functional notation F(a, b), where ‘a’ and ‘b’ stand for the strings in question, the first string and the second string, aka variables. So we might have F(‘hat’, ‘hatter’) with the variable a having taken the value ‘hat’ and the variable b having taken the value ‘hatter’.

The purpose of this function is to enable the activity of one node in our model to jump or transfer to other, similar nodes. More similarity of names of nodes, then more jumping. One, as it happens, of a small number of ways in which such jumping can take place.

The purpose of this post is to say something about how surprisingly complicated the business of similarity can get. And jumping ahead, to conclude with the idea that there is no natural, simple, universal way of doing this. It all depends on the work that you want your similarity function to do.

We start with the rather odd idea that F(a, a) should exist. That the formula for saying how similar, or not, ‘cat’ is to ‘hyena’ should also be able to say how similar ‘cat’ is to itself. Or how similar ‘antidisestablishmentarian’ is to the near miss ‘antidisestablishmentarians’. We then say that F(a, a) should be relatively large, probably bigger than F(a, b) or F(b, a) for any values of a and b that you care to choose. The string ‘cat’ should be more similar to itself than to ‘thecatsatonthemat’ or to ‘the cat sat on the mat’, and certainly more similar, to itself than to ‘hyena’.

Then we say that F(a, a) should be larger than F(b, b) if a is longer than b. Having two copies, or two near copies, of the string ‘wetherspoon’ is more significant, is less likely to be the result of chance, than having two copies, or two near copies, of the string ‘mug’. This requirement allows defining similarity as the number of matching characters, but not as the proportion of matching characters.

Then if string a starts with string b, we want a to be similar to b. So ‘cats’ is similar to ‘cat’. If string a ends with string b, we want a to be similar to b. So ‘unwell’ is similar to ‘well’. And lastly, if string a contains string b, we want a to be similar to b. So ‘antidisestablishmentarian’ is similar to ‘establishment’. Perhaps, other things being equal, with descending similarity as we go through the three options.

We wondered about whether the function F should be symmetric, whether F(a, b) should be the same as F(b, a) for any two strings a and b that one cares to choose. And we decided that we do not want this. We want the node ‘apple tree’ to stimulate the node ‘tree’ more than the other way around. This because while ‘apple tree’ is an instance of ‘tree’, a more specialised version of ‘tree’, ‘tree’ is not an instance of ‘apple tree’. Put another way, all apple trees are trees, but by means do we have it that all trees are apple trees. There are pear trees, cherry trees and all sorts of other trees. Put another way again, we want to have it that if a is a substring of b, then b is should be more similar to a than a is to b.

And we want the function F to be sensible about near misses. To do something sensible about the difference between upper case letters and lower case letters and to do something sensible about possible duplications. To do something sensible about other misspellings. Perhaps, even, to know something about which letters are near which other letters, from the point of view of phonetics or acoustics.

And what about the triangle inequality which says that the sum of two sides of a triangle is always greater than or equal to the third? An inequality from which all kinds of good things flow. An inequality which can be given meaning in this context by saying that the distance between two nodes is some kind of inverse of their similarity, one goes up when the other goes down.

We note in passing that we are working in a language, English, which has an alphabet. In the absence of an alphabet one can still compare sentences by comparing the constituent words, but this is a bit coarsely grained compared with character based comparisons. Perhaps in the case that one does not have an alphabet, much greater reliance has to be placed on how words sound – to the exclusion of words which cannot be sounded (‘cat1b’), which might otherwise be useful – and replacing the problem of misspellings with those of poor or varying pronunciation and poor or varying reception. From the point of view of transmission at least, letters are much more robust than sounds.

Our routine

All of which sounds terribly reasonable. So we try to come up with a function F in VB which does this sort of thing. Or, more precisely, a subroutine.

A first shot is the simple rule mentioned above, to say that similarity is the number of characters that two strings have in common. So the score for ‘cat’ and ‘at’ is two.

We might then add a rule about duplicates, perhaps dropping duplicates from both strings, perhaps scoring the number of matches. So the score for ‘caaaat’ and ‘caat’ might be four.

Pushing a bit harder, do we have the same score for ‘CaAAaT’? Do we have the same score for the pair ‘cat sat on the mat’ and ‘cat sat’ as for the pair ‘cat sat on the mat’ and ‘catsat’? Is the loss of the space from this last string significant? Then switching to functional notation, is F(‘cat12’, ‘cat13’) greater than F(‘cat12’, ‘cat1b’)? Is F(‘red cat’, ‘cat’) the same as  F(‘cat red’, ‘cat’)?

One way to proceed is to try to find a map which tries to map, or match, all the characters of the second string onto characters of the first string. One then scores that map. One then tries to find the map, out of all the possible maps, which gives the best possible score. In this, we might have a rule which says that different characters in the second string must not be mapped onto the same character of the first string. We might have another rule which says that the matching characters have to occur in the same order in the two strings, perhaps with a bonus being given for matching characters which are consecutive in both strings. With a problem here being that there may well be more than one way to do matching of this sort, with one way not giving the same result as another. Perhaps we deal with that one by saying that our similarity function takes the best possible match, the one that gives the most similarity, and leave the work that computation of that function might entail to someone else.
Following the earlier thought, we then need to check what happens with substrings. It would be good if when a is a substring of b, then F(b, c) is always less than or equal to F(a, c) – because if a is smaller than b, we have less stuff in the first string to be matched with stuff from the second – and F(c, b) is always less than or equal to F(c, a) – because we have more stuff in the second string to be matched with stuff from the first.

After some days of this, we wind up with a routine, about one fifth of which is illustrated above. A routine which involves nine constants (for example ‘c081’), all of which need to be given sensible values, and all kinds of choices about how exactly the algorithm is going to work. With our hopefully having made sensible choices, taking into proper account the sort of strings my routine is going to encounter. All of which now needs to be tested on a good sample of same.

Other peoples’ routines

Reference 2 points to a variety of other ways of doing this sort of thing, some of which look to be well established and some of which look a good deal more complicated to compute than the algorithm we have come up with.

But some of them looked nice and simple, with one such being based on the scalar product of the two strings. Unfortunately, that one stopped being nice and simple once one started to say what one meant by the scalar product of two character strings, this product being more usually defined on vectors of numbers rather than vectors of characters.

However, we did like the look of a Russian distance, the Levenshtein distance, described in more detail at reference 3, with the idea being to count how many changes your have to make to get from one string to the other, with the implementation of this distance being a lot shorter and neater than we had expected. A model of concision – if not of comprehensibility, although Wikipedia does include some helpful examples.

A bit of translation work into Excel VB and this new routine is now working, with a number of extra twiddles to deal with things like duplicates and upper and lower case - and with a modest six constants rather than the nine we had before. But not quite as short and neat as at first appeared.

And the first observation is that it seems to give much the same results. Strings which are near each other get good scores, strings which are not get bad scores. Maybe that is good enough. Or do we need to make a more serious comparison?

What about the brain?

The brain seems to be quite good at assessing how similar things are, be those things birds, bees or words. And given that words arrived on the scene quite late, it seems unlikely that any of the brain’s machinery is particularly adapted for the comparison of strings.

Contrariwise, it does seem quite likely that there are places in the brain for particular words, places which can be connected together, rather in the way that we want to connect our nodes together. Places which can be linked together by synaptic connections. With the individual links being directed, and with a collection of links possibly being in one direction or the other or, more likely, something in between, rather like our similarity function.

But we do not know whether places for words are linked together according to how they are spelt, to how they are heard, to how they are said or simply according to how they are used. Or, indeed, whether the building and breaking of links uses all three mechanisms. Remembering in this that hearing and saying are old in evolutionary terms compared with writing.

In any event, while a neural network in a brain might well do something very like the similarity described above, it seems very unlikely that it would start with rules and then code them up. Much more likely that some kind of assembly of neurons implements something or other, then evolves, quite quickly, as the host grows up, evolves towards something which works. Much more likely still that real neural networks are not terribly modular and functions like similarity are all mixed up with lots of other stuff and hard to untangle. Perhaps implemented in different ways in all the different places where some kind of a similarity function is needed.

Turning to genes, it is clear that they specify a general purpose computing facility which somehow has to learn how to do particular things, like similarity, for itself - noting in passing that these same genes have provided the vertebrate retina with the fairly specialised equipment needed to detect useful stuff out in the real world. John Holland, an eminent computer scientist who took an interest in such things, was apt to talk about the Kreb’s cycle for generating the energy needed to keep organisms alive, a cycle which is strong conserved, present in more all less all the larger living organisms that have been around for the last billion years or so. But we are not sure that the fact that a reasonably complicated chemical cycle occurs everywhere, is more or less the same everywhere, means that the same will be true of similarity algorithms. While the genes might have resulted, in time, in assemblies of neurons that do similarity, they do not code for similarity in the ordinary, computing sense of the word.

Closing thoughts

The first thought is that it is surprising how quickly all this can become rather complicated. Perhaps it would all be a lot easier if we used an artificial neural network rather than a conventionally coded algorithm, with all its complicated rules and regulations. We just give our network some indications about how our similarity function is to behave, perhaps by giving it some examples, and it works away, by itself, to come up with something, something which works even if we do not understand the detail of how it works. We might have trained the network, but don’t know what the rules it has come to along the way. Which can be annoying.

Another line would be to write a similarity function with lots of parameters, then to search the space defined by those parameters for an optimum solution, with optimum perhaps being defined in terms of performance on some set of examples.

But a big part of the answer seems to be that it all depends. There is no one way to do similarity and while, other things being equal, it is best to keep things simple, it all depends on what work you want to your similarity to do and what sort of things you want your similarity function to compare. Are we writing a text retrieval system or are we trying to eliminate misspellings? Or what? Or, as they used to teach us in the days of structured methods, don’t start by rushing into code. Start by being clear about the requirement.

The closing thought is that a similarity function only has to be good enough. While it is nice to have short and neat code, that should not be an end in itself. And while it is nice for the code to give the right answer for kinds of perverse examples, that should not be an end it itself either. It is enough to work on sensible, real world examples, most of the time.

PS: some of us used to add to requirements analysis the proviso that the requirement interacts with what it is possible; we are in a feed-back, not a feed-forward situation. Which makes things much more complicated.

References

Reference 1: Before You Know It: The Unconscious Reasons We Do What We Do - John Bargh – 2017.

Reference 2: https://en.wikipedia.org/wiki/String_metric.

Reference 3: https://en.wikipedia.org/wiki/Levenshtein_distance.

Trolley 120

Three trolleys captured from the stash mentioned in the last post. Litter inside consigned to the bin and the three of them tethered together with my very own trolley jockey's strap, perhaps the second time I have had occasion to use it since picking it up getting on for a year ago now. See reference 1.

Tethered, the trolleys were surprisingly easy to walk on flat pavements, much harder going when they got bumpy, with the bumped blind stones being particularly hard on the wrists. Also surprising how much the wind could blow them about; if one did not pay attention one could easily veer into a parked car, doing visible damage.

No wheel locks. Not sure that I would have risked taking three if there had been.

Scored as one trolley as all picked up from the same place.

PS: noticed that the wine bar in Ewell Village has closed, after a year or so. To me it looked a bit too big and basic for a wine bar, so I am not sure what sort of trade they did - and I don't think the twits at reference 2 were either. Certainly not much during the day when I was around. Also noticed the first rat of the year in Blenheim Road, on the home stretch, as it were.

Reference 1: http://psmv3.blogspot.co.uk/2017/04/train-spotters.html.

Reference 2: https://twitter.com/WildsLounge.

A stash

A stash of trolleys from Sainsbury's outside the accommodation block for students from our University of Creation in East Street. Maybe half a mile from the shop, so a reasonable distance to wheel a loaded trolley.

Is it an installation or some kind of a project? Are they making an important statement about the throw-away culture of our consumerist society? Or perhaps, following the closure of both the nearby public houses (the Common Room and the King's Arms), the students are reduced to alcohol fuelled orgies in their bed-sits, for which they need to obtain supplies? Or if I were to visit late at night, would I find of long line of them wheeling each other in trolleys down the centre of the road, while they all chanted some political (or at least polemical) ditty?

Reference 1: http://www.uca.ac.uk/.

Maigret et les Vieillards

This being the fourth and last story of Volume XXI. A story mainly concerned with old people with particles in their names (sometimes called ci-devants), an opportunity for Simenon, once again, to sport a bit of class envy. But also an opportunity to sport some interesting factlets and vocabulary, some of which I share in what follows.

The box tree gets two ecclesiastical mentions. Firstly, a sprig of box was put in the stoup which came with a domestic crucifix. Unable to find out what this was about. Secondly, a parson took his boxwood snuff box out from underneath his cassock. Presumably, boxwood, being hard and fine-grained, was very suitable for making small boxes.

If you were one of the chaps in France, you went to school at Stanislaus in Paris. Large, Catholic and private.

If you went in for diplomacy, you were in La Carrière. Oddly, a usage to be found in Larousse but not in Littré.

While if you fancied the army, you would probably try for the Cadre noir de Saumur. At the time Simenon was writing, a cavalry training establishment. Maybe a bit like getting into the Blues & Royals. Now mainly taken over by civilians.

A marriage blanc in France is one which is not consummated. But you need to be careful with this one because in Iran it means a marriage which has been consummated but which has not been licensed by the ecclesiastical authorities.

If a group of people 'se donner le mot' they are cooking up a story between them. In this case for the police.

The expression 'd'ores et déjà' is a long winded way of saying already, that is to say 'déjà'. Littré talks of 'ores' being an obsolete version of the particle 'or' but I don't see how that helps.

And if you are frileux you are someone who feels the cold.

While a 'vente aux enchères' is the French way of doing auctions. And their cheques have talons, that is to say heels, rather than stubs.

Lots of people in Maigret's world, and presumably in Simenon's world, seem to take sleeping pills. Perhaps the prescription drug which was fashionable at that time, before all today's prescription opiates had been invented.

However, what the story is really about is one of the chaps who commits suicide when he is getting on for eighty, for reasons to do with his love life. His devoted servant, a lady of a similar age with whom he used to sleep on an occasional basis, then shoots him a few more times to cover up the suicide so that he can have a Christian burial.

But the priest - the chap with the snuff box mentioned above, explains that all kinds of things can happen in the split second between pulling the trigger and being dead and that all things considered the church would probably have given the dead chap the benefit of the doubt anyway. He quite probably repented in that split second. No need for a cover up.

Not clear whether he would have taken the same line if he had been dealing with some low-life from Belleville, rather one of the chaps.

In any event, as far as Maigret was concerned, the servant was free to go home. No sign of any charge of attempting to pervert the course of justice or of wasting police time.

Reference 1: https://www.stanislas.fr/.

Reference 2: http://www.ifce.fr/cadre-noir/.

A light lentil

A lentil soup for lunch today, for one.

Two pints of water for four ounces of red lentils. Add a few tablespoons of finely chopped left over potato and carrot. Bring to the boil, simmer for a few minutes. Leave to stand for a few hours.

Bring back to the boil, add three stalks of celery. Simmer for a while.

Meanwhile slice two and a half onions in the way of the segments of an orange and fry gently in a little butter. Chop three rashers of bacon and add them.

When the onions and bacon are cooked, blend them with the lentils and serve with brown bread. At the watery end of the spectrum, with a modest number of dead flies, but entirely suitable for the lighter lunch. For an early mention of dead flies see reference 3.

It was the first time that I had used the new frying pan noticed at reference 1. Entirely satisfactory.

The bacon was described as Sainsbury's smoked streaky, carefully selected and trimmed by hand. I dare say an aggressive barrister armed with the trade descriptions act of 1968, as updated by the consumer rights act of 2015, could have had some fun with it. See reference 2.

Reference 1: http://psmv3.blogspot.co.uk/2018/01/aga.html.

Reference 2: http://www.legislation.gov.uk/ukpga/2015/15/section/11/enacted.

Reference 3: http://pumpkinstrokemarrow.blogspot.co.uk/search?q=hedda+gabler.

Irritation

I have just spent what seems like half an hour trying to book some theatre tickets on the ATG website, an outfit which seems to control a large chunk of London's theatreland.

Hard to find one's way about. And when one has finally succeeded in selecting some tickets, even harder to pay for them.

Then, after I thought that I had finally succeeded in paying for them, the system appeared to freeze.

There was no telephone number visible, but Bing turned one up for me, for a call centre in Woking. So next stop, the call centre - which first explained to me that once I had clicked the pay button, there was no going back, no returns and no refunds. Which they might, from experience, have found necessary, but it is hardly the way to win friends and influence people. It then rang in a rather uncertain way for a while, then put me in a queue (with music) for a while, then cut out.

I have probably not bought anything as I have had no email confirmation - and I am not inclined to try any harder. Moment passed. But given that it is not as if I am that short of the odd half hour, slightly surprising how irritated and cross I had become.

Don't recall having these sorts of problems before, so perhaps their computer was having an off day.

Reference 1: http://www.atgtickets.com/.

Blowbag

It was many years ago that I first read Aldous Huxley's 'Antic Hay', one of his first best sellers and published in 1923. At that time I thought it a tremendous book and remembered ever afterwards that part of the story which concerned inflatable rubber pouches worked into the seats of trousers so that one could sit in otherwise uncomfortable places, perhaps in the seats of a lower grade theatre or concert hall. Known in the book as Gumbril's Patent Small Clothes, an invention which was to propel the young Gumbril out of teaching in a public school, an occupation to which he was not suited. A public school with a chapel with very uncomfortable seats for the staff, who were apt to be confined in them for protracted periods.

So I was pleased, quite by chance, to come across the gadget illustrated. Not built into trousers, but otherwise the same sort of idea. One places it on the uncomfortable place, sits on it and then blows it up until one reaches comfort. It seems to work OK at home, but I have yet to try it out in the field, as it were.

In the meantime, I thought I would take another look at 'Antic Hay', and with my own copy having long been culled, reserved it from the library. Surrey Libraries being quite efficient in such matters, it soon turned up at Epsom Library, having been transferred from some other library in the deep south.

Sadly, although the prose still sparkled, the book had lost much of its appeal, and I failed to muster enough interest in the bright young things of Huxley's circle in the aftermath of the first world war (from which he had been barred by his very bad eyesight). A circle which included luminaries such as D.H. Lawrence. But not the Edward Garnett of reference 1, with the index of Sybil Bedford's biography of Huxley including just one entry for David Garnett, the son of Edward. Perhaps Edward was too busy nursing young authors who needed nursing to bother with quickly successful authors like Huxley. Perhaps Huxley was not considered to be a serious writer.

Taking a further look this morning, I was tempted to give it another go, but in the end decided against. Back to the library with it.

As it happened, while all this was going on, in the course of the expedition noticed at reference 2, the Raynes Park platform library yielded Waugh's 'Put Out More Flags', another comic novel concerning a rather different set of bright young things, published in 1942, just about twenty years later than 'Antic Hay' with this particular copy being a cheap war time edition from 1943. But it was the same story. The prose may not have sparkled, but I failed to muster enough interest. Into the bin with it.

I shall, however, retain more or less unread, my copy of Marlowe's 'Edward II', from which the title of the first book was taken. Line 60, Scene I, Act I: 'Shall with their goat feet dance an antic hay', with the antic hay being a sort of dance already old fashioned at the time Marlowe wrote the lines - which this morning look to be a very apposite place from which to have taken the title. A factlet first noticed about eight years ago at reference 3.

PS: I had completely forgotten that my copy of 'Edward II' had been borrowed from the Wetherspoon's library at Tooting. It seems a pity to return it; it seems unlikely that anyone there these days is going to want it on their shelves more than I do.

Reference 1: http://psmv3.blogspot.co.uk/2018/01/midwife.html.

Reference 2: http://psmv3.blogspot.co.uk/2018/01/rossopomodoro.html.

Reference 3: http://pumpkinstrokemarrow.blogspot.co.uk/search?q=turin+garrett.

Sandbag

A new-to-me road mender's sand bag, presumably intended to stop temporary road signs, the hinged sort which stand on the pavement, from blowing a way. With the two bags, only one of which can be seen in this snap, separated by a sturdy handle.

Altogether a more sophisticated object that the small sack of sand that one usually sees. And surprisingly heavy.

The telephone had trouble with the focus, with the white letters only being in focus in about one snap in two. It seems to be able to get the focus, particularly if you prompt it by tapping the screen or moving the telephone, but it seems to have trouble holding it. I am sure that it is worse now than it was when it was new - and the funny gear slipping noises are certainly new.

Recovered from a heap of spoil in Blenheim Road, outside the joinery factory there, and snapped on the wall at the corner of Blenheim Road and Longmead Road.

Now sitting on the table outside the kitchen window, awaiting deployment.

Trolley 119

Trolley 119 came from the other side of the Sainsbury's car site, roughly gmaps 51.343781, -0.258341. Complete with deployed wheel lock and most of a packet of Pringles. It would probably have been safe to eat the Pringles but I thought that the birds on our back lawn were more deserving.

A wood pigeon is working hard to finish them off as I type. While the lurking cat has not had any kills.

Trolley 118a and trolley 118b

In the event, the trolley mentioned in connection with trolley 117 had become two trolleys this morning. Taking into account the facts that while the trolleys were only a few yards outside Sainsbury's perimeter but were not visible from within that perimeter, I decided that the two trolleys should count as one. Both at near enough gmaps 51.339506, -0.257138.

And having made the decision, did not revoke it when the wheel lock on one of them engaged on the way, necessitating two trips rather than just the one.

And on the way passed a novel use of a shopping trolley, with a maintenance man using one as a saw bench, rather like a workmate from Stanley. Wrong height for using a hand saw but it looked OK for a hand held electric saw of some kind.

And past him, I was rather surprised to find the owner of an Audi soft-top, which I would have thought rather an expensive car, with A5 cabriolets starting from £38,000 or so, letting the lads from Halfords have a go at his soft-top, presumably not working. The sort of thing which I would have thought better attended to by a warranty-preserving dealer, or at least some other kind of expert.

Bare hands

Something has happened to the lids on the bottles of Villa Maria's sauvignon blanc 2017, in that I can no longer get them off, first time around, with my bare hands. Or indeed, with my hands wrapped in a bit of thin cloth, say a tea towel, an expedient which sometimes works with jam jars. Not caring to visit the plumbing department in the garage, I am reduced to using a pair of nut crackers, the larger aperture of which is spot on.

I have not noticed the baristas at Wetherspoon's having the same problem, so maybe they are on some previous vintage. Or maybe they shift enough of the stuff to get a special deal from the Villa Maria people.

Trolley 117

Another Sainsbury's trolley, recovered at the junction of Hook Road and East Street, from in front of the advertising hoarding there.

It was raining, as can be seen from the bit of umbrella top left, which meant that the snaps were taken without the full benefit of two hands, with the result that the camera angle and focus were even more wobbly than usual. So the choice for present purposes rather more restricted than it should have been.

It seemed quite hard work wheeling this trolley one handed, quite heavy on the wrist, so I declined a second trolley on the way out of Kiln Lane. Maybe I will look in again if I go past today - with today being Monday here, although not for followers of PST, that is to say Pacific Standard Time, for whom it is just past 2300 on Sunday.

I dare say if I gave it some proper thought I could work out why logging blogposts to local time would not work. Presumably something to do with their being visible worldwide and the need to manage replication over the various data centres from which blogs are displayed.