“The Constitution does not say that a person can yell ‘wolf’ in a crowded theater.”

— Nancy Pelosi, supa-genius

Dear baking bloggers,

When discussing the Japanese dough-making method popularized by Taiwanese cookbook author Yvonne Chen, please do not write things like “In Japanese, tangzhong means either a warm or thin starchy (flour-based) starter”.

No, it doesn’t. In Japanese, “tangzhong” doesn’t mean anything at all, because the word 湯種 is pronounced “yudane” (and literally reads “hot water seed”). Anything you read or write about tangzhong is based on a Chinese cookbook that adapted a Japanese technique. If you want to call something “Hokkaido Milk Bread”, don’t describe it using Chinese words.

To find specifically-Japanese breads and recipes, search for “湯だね食パン” or “湯種食パン”.


So it looks like it all traces back to blogger and cookbook author Christine Ho, who introduced tangzhong to English-speaking audiences early in 2010, and has published a number of additional recipes, including the one that everyone copied, Hokkaido Milk Toast.

The Chinese name for Chen’s recipe, 北海道牛奶麵包, is indeed “Hokkaido Milk Bread”. It looks like Ho eventually settled on “water roux” as the translation for 湯種, but she’s quite clear that this is a popular Chinese method based on an apparently-patented technique developed in Japan.

Amusingly, the commercial bakery who owns the patent is headquartered in Nagoya, and doesn’t have a big presence in Hokkaido. In other words, it’s possible that neither tangzhong nor yudane is how Hokkaido-style bread is actually made in Hokkaido…

Updated Update (5/19)

I haven’t found a Japanese reference that predates Chen’s book, so unless I stumble across the patent, I’m going to assume that she coined the name 湯種.

After reviewing a number of recipes, I’m convinced that the commercial Japanese method was quite different, and Chen deserves the credit for making it faster and more convenient. Japanese recipes for yudane bread may use either, but tangzhong bread almost always uses Chen’s method.

For instance, this Japanese recipe from a flour company has you mix 100 grams of bread flour with 73 grams of 90°C water, wrap it tightly, and let it rest for 10 hours in the fridge.

This recipe makes the starter a bit differently, having you mix 50 grams of flour, 5 grams of sugar, 5 grams of salt, and 100 grams of boiling water, wrap it tightly, and let it rest in the fridge for 12 hours.

Compare to Yippee’s recipe based on Chen’s book (which predates Ho’s version, but doesn’t seem to have been the (coughcough) starter for the trend in English-speaking baking), which uses a 5:1 ratio of water to flour, and cooks it to 65°C, either on the stovetop or in the microwave (about 90 seconds either way). Where most recipes subtract the tangzhong amounts from the total used for the bread, this one makes enough for two loaves, with instructions on saving the rest for later. Her primary change to Chen’s recipe is adding a sponge.

Now for some fun! Here’s a Japanese translation of an english-speaking Egyptian’s version of King Arthur Flour’s adaptation of Chen’s method, applied to Japanese-style dinner rolls.

I haven’t tried the overnight yudane method, but this morning I made the King Arthur classic sandwich bread using the 6:1 (when converted from volume to weight) tangzhong option in the notes, which cooks it on the stovetop. It’s almost all gone now, so you could call it a success.

(the recipe worked perfectly in my Zojirushi Virtuoso bread machine, by the way, which I bought recently when my old Oster got a little flaky. Since by default it has a 30-minute rest period at the start, I didn’t even bother to cool the tangzhong to room temperature; I just put it in the bottom of the pan with the other liquid ingredients, poured the dry stuff on top, and made a well for the yeast, as usual)

I had odd flashbacks to my childhood, though, because the last time I cooked flour and water into a paste on the stovetop, I was making papier-mâché.

Oh, and the folks at America’s Test Kitchen included their take on Japanese Milk Bread in their book Bread Illustrated, using the same tangzhong ratio as King Arthur Flour. Theirs adds an egg for more richness.

I went somewhere else with this…


Dear Adobe,

I just got email touting the new “Adobe XD CC” app, which promises:

“The future of experience design. No experience required.”

This sounds like exactly what happened when Apple set fire to their decades of usability studies and sucked on a pistol Boldly Invented New Paradigms.

The Penguin is a Harsh Ebook

Apparently the rights have finally been sorted out, and last week Penguin released Heinlein’s The Moon Is A Harsh Mistress for Kindle. TANSTAAFL applies, but they kept it under $10.

3D Cheesecake 2

Consolidating all my files onto the new NAS led me to find a variety of non-animated cheesecake. I don’t have names for all of these, but I’ll fill them in as I come across them.


Dear Wordscapes,

I don’t understand why you’re not accepting this 7-letter word…

Oh, right, ‘resource forks’

The last bit of work involved in merging all my personal Mac files back together on the new 12-inch MacBook was FontExplorer X and all of my fonts. For the past few years, I just grabbed a few at a time from the old machine when I needed them (first and foremost being my improved version of Anonymous Pro for Terminal use), but it was time to get them all back.

Except that when I was done, there weren’t as many as there should have been. I opened up a few old Illustrator files, and got warnings about missing fonts.

Why? Because old Mac font suitcases were stored in the magic Mac resource fork, and somewhere in the various iterations of copying the font directories around, I lost all the resource forks, leaving a bunch of zero-length files. Fortunately, I not only made a disk image of the old Togetsukyō disk, I still had the actual disk on a shelf, right next to the two full backups I made with SuperDuper! when I pulled it from the dead machine.

After that, it was trivial to merge over only the files that were zero-length on the new disk. cp handles resource forks just fine, so I massaged the output of find into a quick shell script, and all was well.

Side note: I have way too many fonts, and that’s not even counting the old shovelware CDs that came with CorelDRAW! back in the day (including the early one they got their asses sued off over).

Oh, dear, I found more fonts…

Just remembered that I saw a Fonts directory on one of the really old external drives I copied to my new NAS. When I opened it, I found another gigabyte of old fonts. This one does include the old CD4 fonts, plus my old Type On Call purchases and everything that was bundled with Illustrator 7 and 9. I’m sure there’s a lot of overlap with what I’ve already got loaded, but best to be sure.

Okay, I don’t need those fonts…

The CorelDRAW 4 fonts have some novelty value (this was the set of Bitstream fonts they licensed after the pirated fonts in CD3 got them sued), but the 26-year-old font archive that predates my arrival in California looks like a steaming heap of shovelware. I don’t even remember where they came from originally, and I already have the DynaCom shovelware CD if I wanted ripoff fonts without decent kerning tables.

And I’ll just delete the disk image labeled “Font Folio 8”. However I acquired it back in the Nineties, it wasn’t by purchasing the product personally, and I can get legit modern versions of pretty much everything on it from Typekit as part of my Adobe Creative Cloud subscription. (besides, it’s more resource-fork suitcase fonts…)

Dear Apple,

So, what exactly is the point of reporting 2^63 - N as the number of free inodes in df output on a Mac? You could at least abbreviate this nonsense in scientific notation, if you can’t report a meaningful number.

“Need a clue, take a clue,
 got a clue, leave a clue”