On Pattern and Rhythm in the Novel

In Aspects of the Novel Forster devotes a lecture to pattern and rhythm. By way of caveat he apologizes in advance for vagueness: "we will borrow from painting first and call it the pattern. Later we will borrow from music and call it rhythm."

His first example of pattern is an hourglass: a shape easily comprehended by the eye and admired for its symmetries, to be found in plots which go neatly from top to bottom and then upend themselves with satisfaction. His second example is a quadrille, a dance made up of chains of cause and effect, where the symmetries are dynamic, rotational, rather than static.

Rhythm is the last aspect of the novel Forster chose to treat, and in this, his eighth lecture, the last before conclusion, I sense that he was growing tired. Again he described two forms. The simple form has a meter that easily comprehended by the ear; it is less a trope than a simple rhetorical device on the order of anaphora, on the order of epistrophe. The more complex form of rhythm is a subtle thing that arises from the action, but can only be understood after the piece has been played and the orchestra silenced, the better for the listener to fit the parts together in his mind and grasp the whole. It is like Pointillism, which also requires the audience to step back.

Unusually, Forster ended his lecture without relating pattern to rhythm. If he were writing today with more time and more energy, I would like to think he would have ended the lecture thus:

If pattern is an hourglass, then rhythm is the sound of sand grains falling. Form follows function: the shape of the hourglass follows the sand. As the sea is in the fish and the fish is in the sea, the hourglass is in the sand and the sand is in the hourglass. Without sand the hourglass is useless; without hourglass the sand is dead. Together they live: they measure out time as a novel measures out story.

Within a tiny seed, goes the saying, lies a mighty oak. And so it is with the second form of rhythm: it provides a motif, an inspiration, a shard of a hologram, from which the whole work can be deduced. Show any competent designer a clepsydra, then say "make it dry, portable, and reusable"; come back in a month and he will necessarily give you an hourglass. It is an obligate consequence of the requirements. Show a novelist a fragment of rhythm; come back in a year and he will necessarily give you the pattern.

Letter to SG Perm Secs

I was invited to lunch earlier this month with a couple of Singapore Permanent Secretaries, for Manpower and for Public Service under the PMO. We had a great time chatting and I reaffirmed my positive impression of the calibre of Singapore's public sector leadership.

Today, I finally got round to sending them a thank-you letter:

Thank you very much for inviting me to lunch. I enjoyed our conversation thoroughly. Lionel has invited me to speak at the CSC and I look forward to the opportunity.

I had a few thoughts following from our discussion. May I share them?


We observed at lunch that our fertility rate is now 1.22.

One of my friends, a born-and-bred white-collar Singaporean in her early thirties, married with no children and no intention of having them, commented that some species just don't breed in captivity. And when I flew in to Changi the other night, I drove past HDB flats with fluorescent lights blazing in every corridor even at 1am, and I thought to myself: don't they keep hens under lights, too, to maximize egg productivity?

Or maybe the birth rate would go up if MDA stopped blocking, who knows. :)


More seriously though, a low birthrate is currently conceived as a problem. But why not reframe it as an opportunity? One of the chief causes of entrepreneurial risk aversion is the need to support a mortgage and family. But if Gen X and Gen Y are opting out of the gene pool, let's respect that choice; maybe they're not failed breeders: instead they're latent entrepreneurs and investors. All the money that would ordinarily be budgeted on children can go towards entrepreneurship. SDU's loss could be EDB's gain!

There is a larger issue here. We live on an overpopulated planet, and sooner or later we will have to find sustainable economic models that are not premised on population growth.


In the same way that China and the US hope to reap economic benefits from embracing clean energy, I believe Singapore can pioneer a model of sustainable economic growth that aligns with zero population growth – not necessarily here, but globally. If the native population with their excellent educations and economic stressors do not reproduce, we bring in immigrants to make up the difference. Sure, the high-end immigrants tend not to reproduce either. But is that really a problem? The world is a big place, and we can keep skimming the cream for a very long time. In fact it's cheaper: they come already educated and start paying taxes right away.

I was re-reading Richard Florida the other day and Singapore has done a lot of things right since his first book came out.

The irony of anti-immigrant sentiment is obvious: the objectors wouldn't be here but for immigration. I would be perfectly happy to see the new face of Singapore look whiter, browner, or yellower than it is today. Would we rather have a meritocracy or a bumiputra policy? It's a tough choice, but I trust our leadership to make it. The world is flat and survival today demands the same kind of tough-mindedness as the first generation displayed fifty years ago. The cells in our bodies are constantly being replaced, but shapes of our bodies, and our souls, stay the same...


Speaking of immigration politics: we already have NMPs. What's stopping us from giving PRs and guest workers the right to vote for non-citizen MPs?

Even if they had only a voice but no vote, an NCMP could still represent their constituencies in Parliament, and it would send a strong message to people inside and outside Singapore that we respect foreign talent. Guest workers deserve protection, and PRs deserve representation, especially the ones who have been here for years but haven't taken the final step of citizenship. This is the 21st century and citizenship doesn't need to be all-or-nothing. After a few years, when the novelty has worn off, the lines between guest worker, "local", and foreign talent would be thoroughly blurred, and I think that would be for the better. True, there are some logistical challenges in administering the constituencies, but if Facebook can manage 500 million users flowing in and out of their systems, surely Singapore can manage 5.

(I got the idea for NCMPs from George Friedman's book The Next 100 Years.)


If you'd like to meet the geeks and technopreneurs who I personally consider the new face of Singapore, I'd be happy to show you around one of these evenings. The members come from all over the world: Seattle, New York, and California, but also Austria, South Africa, Cuba, and Hungary. They've all moved to Singapore in search of opportunity and if we want to encourage an equal entrepreneurship among our native talent, the least we can do is level the playing field for our scholars, already disadvantaged by NS, by encouraging them to come back and work for startups, not just big MNCs.

By the way, I've been visiting the local IHLs to encourage the ICT/IDM schools to teach a more startup-friendly curriculum. Many of them have incubators on campus but teach only those programming languages that are suited for big banks, so there's a bit of a disconnect. Yesterday I met with the deputy principals of Ngee Ann Poly and Singapore Poly; next I will meet Temasek and Republic and engage further with their IT faculties and entrepreneurship development departments.

Salmon Sous Vide

One banana per person, cut into rounds and caramelized in browned butter on medium-high heat until fragrant, mushy, and lightly crisped, about 10 minutes.

Top with: one square of sashimi-grade salmon fillet, brined for 10 minutes then rinsed. Sous vide at 108°F for 20–30 minutes (due to thickness). Sprinkle several drops of naturally fermented regular soy sauce.

Top with: goat feta, one cube. Grind some fresh ground black pepper onto the goat feta for visual effect.

Plate amid basic vinaigrette green salad with avocado and mandarin orange wedges, de-seeded.

Note that at 108°F the myocommata remain ungelatinized and can be easily distinguished. The myotomes have a texture variously described as custard-like or between ceviche and lox.

Oxtail Sous Vide

Plan on two large pieces of oxtail per person or three to four small, approximately 500g total; reduce accordingly if other dishes are being served.

Season lightly: anise, bay leaf, garlic powder, salt, pepper, fennel, or what-have-you. Seal. You may wish to double-bag to prevent bones and spices from poking a hole.

Simmer at 57°C / 135°F for 24 hours, followed by 70°C / 158°F for 48 hours.

If you're in a hurry, you may skip the first 24 hours, but in that case try to go for 60h at 70°C.

Open bag. Pour meat juices into pot, reserving the chunks to warm under aluminium foil. Reduce on medium-high optionally with a splash of wine, sherry, or cognac.

If there are vegetarians at the table, simmer juices together with whole roasted garlic cloves, carrot slices, previously browned onions, leeks, or whatever other stewing vegetables are at hand.

Reduce until the sauce is thick. Scrape the bottom of the pot to release fond. You may want to stew the veggies separately if they need more time. Or skip them completely.

Use hot pan to sear the meat chunks to a dark crust. Be careful to maintain structural integrity – the meat will fall off the bone at the least provocation. Use tongs and metal spatula to give maximum support to the meat. Or just use a blowtorch.

Serve on noodles in a soup, dry on mashed potatoes, or solo.

Bulk Variation

I cooked this recipe for 50 people the other night. At catering scale, after the 72 hours are over, prepare two large rice cookers. (Commercial rice cookers range from 5L to 10L capacity and above.) No sous vide controller need be connected to the rice cookers.

Use a colander or strainer to separate the meat into one cooker (on keep warm, lidded) and the juices into another cooker (on cook, uncovered). Reduce the juices, using a spatula every ten to twenty minutes to remove fond. While the juices are reducing, blowtorch the meat chunks. When the juices are sufficiently reduced, after 30 to 120 minutes, optionally blend the juices to emulsify, then pour or ladle over the meat, coating evenly. Hold hot or serve. For maximum presentation, plate oxtail before diner then ladle.

Leftover meat may be deboned, pulled, shredded, and refrigerated in a jar, rillette-style, under a layer of sauce.

Disrupting CaseMap with EverNote and MindNode on OS X

My Court Case

Starting in June 2010 I was involved in a civil case. What about? Money, honey. After a year of trading affidavits and unsuccessful negotiations, we went to full trial in open court. Over the course of six weeks, a total of 16 witnesses took the stand. Four sets of lawyers (five by the time the trial ended), including two senior counsel, plus court hearing costs, added up to over $40,000 a day in expenses. The national newspaper tracked the story and the story even got picked up by a number of random blogs.

The documentary evidence occupied twenty-two volumes of paper folders containing over 2000 documents spanning 7,326 pages.

More than a year later, in August 2011, the trial ended. The judge dismissed all the opponent claims and awarded my side just about everything we asked for, and more.

I used several applications to manage the evidence. The primary tool was Evernote. In a supporting role, we also used Dropbox, Mindnode, TimeLine3D, Adobe InDesign and Adobe Illustrator.

Here's how I used those tools to help cut costs, marshall arguments, optimize my interactions with lawyers, and win the case.

The Incumbent: Casemap

Casemap by LexisNexis has all the charm and effectiveness of a Soviet tank. It manages lawsuits by brute force: every argument or claim is recorded into a hierarchy of issues; every issue is supported by a set of facts; every fact is grounded in a document. Along the way, the dramatis personae go into a persons table. It feels a lot like a hypertrophied MS Access application. The TimeMap extension throws up your case history in a timeline whose graphic design conventions reminded me of an eight grade science fair.

Twelve years after its debut, Casemap finds daily use in large law firms which also have all the charm and effectiveness of the Soviet military bureaucracy. Casemap outputs affidavits the way Photoshop exports JPEGs. It deserves its leading position in the ecosystem: it helps you structure your thinking, marshall evidence, and get new members of a legal team rapidly up to speed.

But as a knowledge management tool it feels like a dinosaur about to topple. I built an 80% approximation to Casemap using a nimble, mostly free team of small mammals.

Tools I Didn't Use: Zotero and MacJournal

First, the tool I didn't use: by all accounts Zotero is pretty awesome for academic researchers, and if the majority of my resources were online rather than actual paper documents, I would probably be using it now.

See also

Fujitsu ScanSnap S1500

Liberate all your paper documents onto the computer.

Begin by scanning one folder at a time. My rule is one PDF per staple. For efficiency I defer the OCR to an overnight batch using Adobe Acrobat Pro, or you can just leave that job to Evernote.

Write onto each paper folder the time you scanned it. Place papers back into the folder in order of time scanned. After uploading the PDFs to Evernote, bulk-tag the scanned files with the name of the folder it came from, so that when you next need to go from soft to hardcopy, you know where to find it.

There's no need to muck with the filenames: an ISO8601 format of YYYYMMDDHHMMSS provides sufficient uniqueness. Sometimes I give each physical folder a corresponding folder under my scans/ directory. The Windows Evernote client knows how to watch subdirectories of scans/ for new material.

When you're going through your physical papers you may be tempted to assess each document for its relevance to your case. Don't. This is a layering violation. Just scan everything and examine them all once they're on the computer. The rule is, either a document is in the system, or it's not; a binder containing some scanned and some unscanned documents must be treated as wholly unscanned. You can deal with duplicates later. Without the confidence that all your documents are in the system you will flounder in anxiety.

Get a Second Monitor

Buy a second monitor. If you can, buy a third. The one advantage paper has over PDF is that you can flip through it fast, spread it out on a table, and pencil in notes. You can become just as fast doing this on the computer, but it takes the right gear and a bit of practice.


Evernote is probably the best general-purpose document-oriented knowledge manager out there today. If you didn't do your own OCR (thanks, Adobe Acrobat Pro!) on the PDFs, Evernote (Premium) will make them searchable anyway (yea, even unto the handwriting). The OS X desktop client has really smooth PDF integration: when you put a PDF into a note, Evernote spreads out every page of the PDF for viewing and for displaying full-text-indexed search results.

Tag each note in at least three ways. Source tags describe where the hardcopy may be found: for example, "blue folder 12A". Content tags describe the content or nature of the document: for example, "financial statements". Argument tags associate the document with the high-level arguments and goals that it supports: for example, "restitution".

You can also tag for authorship: for example, "by Smith".

Now that you've tagged all the documents, you can easily search in a structured manner: "I want to see all the documents that were written by Smith which support my claim for restitution." This becomes useful when, say, you're preparing for Smith's cross-examination.

I use "unimportant" as my catch-all tag, and "rescan" for anything that didn't come through well the first time. To find all documents that have not yet been tagged, use the search filter -tag:*

Retitle your notes with YYYY-MM-DD and some text appropriate to the contents of the note. That datestamp and title will later go into your timeline. When you sort by title, you effectively get a sort by timeline.

The goal in all this is to minimize the number of times you visit a given document. Lawyers have armies of associates tasked to do the following:
    foreach issue in (list of issues):
        foreach document in (all documents):
            if (document is relevant to issue)
                then add a coloured post-it flag.

As any computer scientist will tell you, this is wasteful, because you've got an O(MN) operation with a slow inner loop. If there are ten issues, you loop through all the documents ten times.

The faster approach, obviously, is:
    foreach document in (all documents):
        foreach issue in (list of issues):
            if (document is relevant to issue)
                then tag it with the issue name

Now for the gang aft agley: new issues will arise, and you will have to revisit documents to add tags. But all the other tags you've previously added should reduce the number of documents you have to revisit: you can filter by content tag, or by date, or just do a full text search against the raw document text, and associate issues with documents that come up in your results.

During the discovery phase, your lawyers will collate these documents into one or more sets of disclosures. For each document, you will make a decision: disclose or not? Instruct your lawyers, as they move through the documents, to tag the disclosure status of each document. If they disclose it, tell them to tag it with "disclosed 20110818" – that means it was disclosed in a filing on that date. If they decide not to disclose it, tag it with "undisclosed". This is useful because later on, when reviewing documents, the question will inevitably arise: "did we disclose this?" With the tags, you will know.

In a yet later phase, all the disclosures will be collated into one or more Bundles of Documents, which you will refer to at trial. In my court case, we had a First Agreed Bundle (1AB) and a Second Agreed Bundle (2AB) and so on. Each document is Bates-stamped with a unique page number within that bundle. By the time the Bundles appear, you will likely by up to your neck in AEIC drafting and witness prep, so find a volunteer to go through the Index to the Agreed Bundles (which will come, on paper, from your lawyers) and, for each document in that index, match up the appropriate note in Evernote. As a sanity check, that note should carry a disclosure tag. In any case, add a label to the title of the form "BUNDLE-1AB5-2345". That means it's in the First Agreed Bundle, volume 5, page 2345. That makes it easier for you to go back and forth.

Here's what your notes will end up looking like:
Title: 2011-08-20 BUNDLE-1AB6-1315 email from meng to Smith: You're in my sun, buddy!
Tags: folder green 6B, disclosure 20110401, email, by meng, smith, sunlight

If I want to find the original, i know to look in the green folder 6B.

If I'm wondering if this has been disclosed, I know, because the tag says it was disclosed as part of the 20110401 filing. In fact, I know that it's on page 1315 of volume 6 of the first agreed bundle.

If I'm looking for all documents that have to do with the sunlight dispute, I can do a search for tag:"sunlight". If I know it's an email I'm looking for, and not, say, a balance sheet, I can add to the search tag:"email" as well, as opposed to tag:"financial statements".

And if we're preparing to cross-examine Smith, I can add to that search, tag:smith.


In a lawsuit, one makes arguments supported by principles and by evidence. I'm leaving the principles to the lawyers: my job is to produce facts and the documents which illustrate them.

Some arguments rely on others, so the computer scientist in me automatically prescribes an argument graph whose leaves terminate in documents which contain facts.

MindNode makes it easy to sketch up an argument graph, anticipate rebuttals, and associate relevant evidence. Like most mind mapping software it thinks in terms of trees, but if you co-locate multiple nodes, you can mouse-select them at the same time, so you basically get close enough to a real DAG: a directed acyclic graph. Yay!

You can have multiple trees. Define each of your litigation goals as a root.

Give each node its own tag. Tag each supporting document with the issues that it supports. When you compile a given argument, those tags will become crucial to extracting relevant evidence.

Timelining with a Google Docs Spreadsheet

I'm surprised at the lack of good timelining software for OS X. I remember seeing a nifty Silverlight demo which showed stacks of thumbnails bouncing around and spontaneously reassembling themselves according to different organizational schemes. It's not impossible that a future version of Evernote could map all my notes on the X axis by time (created, updated, or in title) and separate them along the Y axis by tag.

It's theoretically possible to use Timeline3D from Bee Docs to visually represent all the documents in one place. As with Mindnode, the "Link" feature gives you one-click access to the source document, which is great.

Evernote exports notes to XML. I wrote a simple Perl script that parses an ENEX export and produces a tab-separated text file suitable for import into Timeline3D. Voilà, we have a poor man's TimeMap.

Unfortunately Timeline3D's import function is very bare-bones: while the presentation software has support for multiple event rows, the importer doesn't.

I gave up when I found that it was impossible to import multi-track timelines from CSV. Sadly, Bee Docs has the feel of abandonware: really easy features go unadded for months.

In the end, I did two things. I kept a basic timeline in a Google Docs spreadsheet; this tracked major dates and events. And I wrote a timeline-to-SVG tool that drew on both the Google Doc and the Evernote SQLite database. I opened the SVGs in Illustrator and manually placed relevant documents into the canvas.

Adobe Acrobat Pro

With Preview under OS X, you can add annotations easily. But you probably want Acrobat Pro for the OCR engine.


This is getting to be a little over-the-top, but you can also use Prezi to give your affidavits and exhibits a sort of semantic-zoom relationship with each other. You will need a Premium account to keep your docs private, though.

I didn't have time to do this.

DropBox and Spotlight

Evernote will sync a complete set of your notes in the cloud. DropBox serves as a backup storage solution.

Evernote full-text indexes all notes for fast search. Spotlight serves as a backup search solution.

In my case, we ended up using DropBox to store the PDFs of the Agreed Bundles as received from opposing counsel, and also for miscellaneous interlawyer correspondence, court orders, and other documents. If you argued that all of this should be in Evernote, I would agree with you, except Evernote has a 50MB file limit and some of the PDFs we got were like 300MB.

How To Compile An Affidavit

Your affidavit makes one or more arguments. Your lawyer (and you) will often want to see all the documents that have to do with a particular argument or set of issues.

Easy! In MindNode, you've already structured your argument tree to indicate which arguments visit which issues.

And you've already tagged, in Evernote, all the documents by issue.

So, whatever the argument is, you can start by making a list of all the issue tags relevant to a given argument subgraph.

In Evernote, filter notes by those desired tags. Sort by note title. Export all those notes and hand them to your lawyer with a ribbon around them.

For extra credit, shove them into Timeline3D if you're the visual type: you can print that timeline and give it to your lawyer as a one-page refresher.

Other Comments

You will run into some extraordinary frustrations when using Evernote. Its lack of support for inter-document hyperlinks is legendary. Skype has its own URI scheme. Evernote should have one too. I predict that if and when Evernote adds WikiWords support it will gain a whole new following.

Living on Cloud Eight-and-a-Half

My mother has just bought a new Macbook Pro 15".

This time, things will be different. This time, her laptop will be a thin client.

(Regular readers, please forgive the occasionally negative tone of this review: a pessimistic perspective is necessary in any risk assessment of infrastructure.)

Thin Client General Principles

Data Recovery. There should be nothing irreplaceable on the laptop. Losing the laptop should lose no data.

Data Security. Should the laptop fall into enemy hands it should disclose no secrets.

Data Replication. Data on the laptop should be available off the laptop.

State Recovery. If the device gets lost, she should be able to reinstantiate it in less than four hours of fingers-on-keyboard work, and be back to where she was before.

Thin Client Specific Requirements

She has about 10 gigs of important data and 20 gigs of photos.

She runs the standard Microsoft Office suite.

She uses some software that is Windows-only.

She cares about private sharing, but not public sharing.

Thin Client Architecture Requirements

Multihoming. The thin client should sync with at least two independent cloud app providers, so that single-provider outage will not affect uploading.

Multi-modal sync. She wants to sync regular files. She also wants to sync email, contacts, calendars, and bookmarks too.

Should work across Windows, Mac, iPad, and iPhone.

Cloud Storage App Providers Compared

For file/document syncing, the candidates are: (as of 20100802)

iDisk, part of MobileMe. Minor drawback: in my experience, iDisk falls out of sync about once or twice a year requiring a complete manual rebuild using rsync, which puts recovery beyond the abilities of most end-users. Major drawback: integration with Time Machine is clunky: TM offers a beautiful UI for unlimited backup versioning of every file on your computer except your iDisk folder. Given Apple's stellar usability reputation, this counts as an embarrassing fail.

DropBox. Great syncing so far: features like LAN sync show that this dev team have the most mojo. Unlimited version support with the "pack rat" feature, which exposes S3's underlying versioning at a significant premium ($40/year). Drawback: obsessed with its special MyDropBox folder: anything inside, or symlinked from, that directory gets synced; anything outside doesn't. For now, workarounds are necessary.

SugarSync. Is able to sync any folder on disk, not just a special subdirectory. Versioning limited to previous five. You can email attachments to SugarSync and they'll automatically show up in a folder.

Evernote. Hews closest to the Weinberger tag-based philosophy. Full-text-indexes all content, notably searchable PDFs. But because it's document oriented, not filesystem oriented, exposing the EverNote repository requires extra work. Being integration whores they have an impressive array of third-party partners on tap.

JungleDisk. Offers direct access to S3, which means you have to sign up for S3 yourself and get keys. Being an AWS developer I find this totally natural, but since they were acquired by Rackspace I'm guessing the original passion is gone, because the OS X app has some weird and fundamental flaws, like not showing me any subdirectories of preexisting S3 buckets, and there's an iPhone but no iPad app. The major pro here is the pay-as-you-go pricing for storage: at $0.15/gig/month, storing 10 gigs would cost me $18 a year, compared to DropBox's $100 and Sugarsync's $50.

For every other modality:

Google. Storing, syncing, and searching of email is second to none. Also provides easily synced contacts and calendars. Google Docs with is an intriguing alternative. Alternatively Google Docs over FUSE with format autoconversion integrating with one of the above methods.

See also SugarSync's comparison matrix.


The ideal cloud storage app would have:
- the full-text-indexing and search capabilities of EverNote.
- the unlimited directory scope of SugarSync.
- the unlimited versioning and LAN sync of DropBox.
- the pass-through pricing of JungleDisk.
- the online editing interfaces of Google Docs.


In the real world, to achieve something like the above, we'll use a combination of:

- DropBox as the primary sync manager across multiple desktops and laptops
- SugarSync as the secondary backup
- Time Machine / Time Capsule as the local tertiary
- Spotlight to handle the full-text indexing

SugarSync is necessary because I don't have direct access to the underlying S3 objects. If I needed to flash a new client at a time when DropBox was down, I could use SugarSync. If DropBox were more like JungleDisk I would be less concerned about DropBox downtime.

The Time Capsule will act as a local fileserver cloud-in-a-box thanks to Remote File Access. All the external drives currently floating around the house will plug in to it. I'd like to see it RAID-1 mirror the internal with the external devices, but I don't know if that's possible. We're also going to use the Time Capsule as a base station, to split the private net from the public guest network, and retire some of the existing network gear. The 2wire will go back to running as just a DSL modem; we'll turn off the NAT and Wifi.

Because Vista and OS X offer usable search, we'll pass on EverNote this time round, though it's intriguing enough that I'll keep playing with it on my laptop.

Because iDisk doesn't work very well compared to DropBox, we'll pass on MobileMe. We don't even need it for contact and calendar syncing because Google does that job very well now. The only thing we lose is Safari bookmark integration, but my mother doesn't need that, and if we did, we could use delicious or pinboard.

We'll do security using FileVault and password timeout locks.

Sous Vide Hamine Eggs

I have just (at 11am) inserted the following eggs in a 71ºC (160ºF) sous vide bath:

- A: 1 egg, in ghee, in ziploc
- B: 1 egg, in ghee, in ziploc
- C: 1 egg, in sodium bicarbonate solution, in ziploc
- D: 1 egg, in ziploc
- E: 1 egg, no shell, in ziploc
- F: 1 egg, no ziploc

This experiment should produce several varieties of eggs hamine, and answer the question: is the browning of the egg white attributed to coffee grounds and onion skins in the traditional recipe due to pigment transfer as in tea eggs or (as I suspect) really because of Maillard reaction of glucose with albumen proteins in an alkaline environment?

Apparently the same browning occurs in antique photos.

Stay tuned for details!

Recipe: Roasted Sous Vide Lamb Shoulder

I'm rocking the restaurant cuts.

Obtain 1 lamb shoulder from MMMM or your local Cryovack-capable butcher.

Instruct them to season with some subset of rosemary, cinnamon, mint, paprika, thyme, parsley, black pepper, garlic powder, salt, etc, or whatever strikes your fancy.  Fold compactly, exposing fat cap, and vacuum-seal.

Sous vide for 12 to 36 hours (ideally 24) at 135ºF (58ºC).

When it comes out of the bath, the meat should already be falling-off-the-bone tender and more succulent than anything else you have eaten this month. You can taste it at this point, and if you are satisfied, go ahead and serve.

To sear: roast in 400ºF (200ºC) oven until fragrant and fat cap is browned, about 10 to 20 minutes.  If your oven has a rotisserie function, now's the time to use it!

(If it falls apart on you before it goes into the oven, then the next time round, chill in ice bath while still in bag to solidify the fats; you can refrigerate for up to a week.)

Oh, and by the way, there's no need to trim the fat out of some vague religious sense of dietary guilt. Saturated fats are good for you! It's the biscuits that you need to avoid. I know this goes against conventional wisdom; if you don't believe me, please read Good Calories, Bad Calories before indulging comment reflex.

How to Dim Halogen Track Lights

A short compendium of things I've discovered in my quest to dim halogen track lighting.

The standard MR16 halogen lamp runs on 12V, "low voltage".

It is possible to dim these lamps. Tungsten halogen is designed to run hot; otherwise vapor deposition on the bulb reduces lifetime; to solve this, run the lamps at full power for a while to promote recovery.

Transformation of 240V to 12V may occur either integrally, within the track head unit, or by way of a remote transformer powering the entire track at 12V.  Remote transformers can be hidden visually and sonically (dimming creates a small amount of noise), and allow one to use line-voltage track heads, which may be fractionally cheaper.

There are two kinds of transformers for 12V halogen: inductive (L) and capacitive (C).  Inductive transformers operate on a magnetic principle, and contain a core-and-coil component.  Capacitive transformers use solid state electronic circuitry.

Depending on the type of transformer you're dealing with, you may have to obtain a leading-edge dimmer (L) or a trailing-edge dimmer (C).  Some high-end "universal" transformers can handle both types of dimming.

Apparently some dimmers are "magnetic", which means they're particularly appropriate for inductive transformers: they carefully chop the leading edge of both phases symmetrically, so there is no net DC flow.

There's a restriction on the cable length between dimmer and transformer because chopping the waveform results in all manner of RF noise, and anything beyond 2m turns your wiring into an antenna.

On the subject of length restriction: 12V track is much more susceptible to resistive loss than 240V track; this is the reason long-distance power lines spin up to very high voltages.  Over long track runs, you might notice that the lamps far away from the transformer are dimmer than the nearer lamps; this is because the copper (or whatever) in the track itself eats power.

Professional industrial applications use addressable dimming.  The prosumer version can be found at