2008-01-12 (Sat) 著作権法違反です [長年日記]

_ [srv][obsd]Stupid Install Tricks: siteXY and install.site (undeadly)

そういえば site42.tgz って試したことなかったなぁ。

_ g11n vs i18n 勝手に英訳しようとして挫折

勝手に https://www.hi-matic.org/diary/?20080110#10-1 を英語っぽくしてみたり。いまだに the の使いかたがわからないなぁ。muradaikan さんに丁寧に教えてもらったのに。

g11n vs i18n

Another (not-so-good) analogy going on here.

You see a green, running man on almost every emergency exit sign. That's a pictogram. Pictograms are a good example of g11n. They are globally understood. On the other hand, the world has not shared one single emergency call number yet. You may have heard that, when the former coach of the Japanese national soccer team fainted, his non-Japanese wife didn't know how to call an ambulance. How helpful a single shared number would have been!

g11n people say,

"Why don't we standardize the emergency number as 911 all over the world?"

Well, the idea of 911-anywhere is very simple and ideal. But it requires millions of numbers to be changed in areas where 911 is already in use as a prefix. Also you need to inform people about the change, and keep both the new and the old numbers co-existing *1 while migrating *2. (In case of migration failure, you will have no emergency number if you have invalidated the old number.)

True, you may be able to share, not only the number but also an emeregency call center. That's efficient.*3 However, you have to design it not to do a poor job. *4 For example, operators should not answer to Japanese caller in English, nor send fire engines from Beijing to a flood in Melbourne. All these smart processes have to be done in quick. Can you still keep it simple? Haven't you dreamed a pipe-dream?

Moreover, you may be annoyed by idiots who call errno's on their cellphone screens like "Failed to transmit your mail (911)." You may have to expect any other unexpected situations. You know, it is an untested new system.

i18n guys say, however,

"Why don't you add an emergency button to all the phones in the world?" In other words, they suggest the new phones with one tiny button to automatically

  • determine the current location and
  • search the number database (Japan/119, USA/911, ...) and
  • report the emergency.

The real number is hidden behind the button. And the good old systems (119 in Japan and 911 in the USA) are kept running as a backend for the new system, though it may defer a real simplification. This idea is in accord with the dont-touch-what-is-working rule. (-because-a-miracle-has-made-it-work)

Nevertheless, it should take a long long time *5 to implement the button on all the phones in the world. Until then, the whole benefit of "the one way anywhere" won't come. In this blank period, g11n people will agitate users shouting "simple is our motto", and indoctrinate most users, I'm afraid.

Moreover, it is possible that the implementation will result in a too complex method to remember ("turn it three times to soften it and release a drop of ...") because the gaps between countries are so large.

Doubt it

Here, although g11n people claim "What a simple world it will be once it is unified! Imagine there's no countries," even then you have to ask,

  • "Is it really simple?"
  • "Have I not underestimated the cost of migration? (Is it possible at all?)"
  • "Won't it unfairly penalize anyone (especially late-comers)?"

to make sure they haven't cheated. Their implementation could also end up being at the same level of complexity or cost-inefficiency as of the i18n approach in the end. And if they believe that keeping legacy thingies is a PITA, they will have to coerce Stalinesque migration!

Also, don't trust i18n guys who make an excuse like "It should take so long but it will be better tomorrow." Don't Believe The Hype (as Public Enemy says). It's possible that their phones won't be released for ever. (Do you have an SED TV?) So "code then spec, not spec then code."

You have to compare the complexity of the wrapper interface, too, with the disadvantage of giving up the legacy one.

*1 just keeping the numbers is not so hard, but creating a Unicode conversion table from/to all of the existing charsets takes very very long time...

*2 the migration can drone on for ever

*3 say good-bye to ISO-2022 and forget about the need of state transition, eh?

*4 if you stupidly choose 16bit for your universal charcode, you will need to invent an ugly idea such as surrogate pairs. also you will need language tags and variation selector for a too tightly unified ideographs...

*5 in fact, C1X regretted that implementations of C90AMD1 and C99 wchar_t API hadn't come sooner

«前の日記(2008-01-11 (Fri)) 最新 次の日記(2008-01-22 (Tue))»