Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.3 4.3bsd-beta 6/6/85; site talcott.UUCP
Path: utzoo!linus!decvax!ucbvax!ucdavis!lll-crg!seismo!harvard!talcott!tmb
From: tmb@talcott.UUCP (Thomas M. Breuel)
Newsgroups: net.internat
Subject: Re: The real work of internationalization
Message-ID: <533@talcott.UUCP>
Date: Fri, 18-Oct-85 10:57:10 EDT
Article-I.D.: talcott.533
Posted: Fri Oct 18 10:57:10 1985
Date-Received: Sun, 20-Oct-85 09:30:42 EDT
Organization: Harvard University
Lines: 72

In article <198@l5.uucp>, gnu@l5.uucp (John Gilmore) writes:
> :-) Of course nobody in China would ever program in Chinese.  They'd just
> learn English because it's the natural language for talking to computers.

I think that our design of programming languages has been influenced
strongly by our natural languages. The question of lexical
analysis doesn't really make sense in Chinese, for example.

Of course you could design a programming language that uses
Chinese characters as its terminal symbols. Given how primitive
the vocabulary of programming languages is, and how unrelated
the 'mnemonic names' are to the real-life meanings of the words,
it is hardly worth it, though. 

> These [special purpose word processors] already exist 
> are are not the subject of this newsgroup.  We're
> talking about extending all the benefits of Unix (I presume you think
> Unix is a nice environment to work and play in, yes?) to people who
> speak and write differently than you do in Murray Hill.

As I posted before, I think it is impossible to implement an internationl
user interface at a low level in UN*X because one of the strengths of
UN*X is that the user interface is identical with an interactive
programming language. If you change the user interface (i.e. change
the name of 'grep' to something else, all shell scripts using 'grep'
will break). Remember what trouble it caused when someone decided that
the extra space in 'date' output was ugly and did away with it?
This reliance upon fixed ouput formats, fixed names, &c is not
bad programming, but a logical consequence of the UN*X philosophy.

Now, that doesn't mean that you couldn't cook up a shell for UN*X
that encodes Chinese characters in ASCII (for file names) and 
translates between Japanese commands and 'English' commands, 
But once you change anything lower level than that, like the names of system
utilities or the output format of almost any program in UN*X, you
can forget about sharing software.

Altogether:

-- I doubt that a hybrid system that understands all character sets,
   all string orders, all national date and time conventions, &c. has
   a chance in the west, because of the overhead and cost involved.
   Maybe a system that can handle just all Roman character set based
   languages has a chance, although I even doubt that...

-- File exchange, program exchange, networking, or any other kind of
   communication between machines with different character sets is
   a nightmare and very likely not to work. Just the difference in
   byte order between the VAX and the PDP is causing lots of grief
   already.

Conclusions:

-- Before you start screwing around with UN*X, please make a backup
   copy so that there is still something working around when you are
   done (anyone have a copy of 4.1 :-). 

-- Of course, in an ideal world, everybody could sit down at his
   terminal, type to the computer in his natural language, and the
   computer would automatically do the rest. Now, I am not opposed
   to that idea, I would just like to hear more reasonable proposals
   of how to do it. And, honestly, I don't think that you can begin
   by hacking namei or by starting to put funny characters into
   your logname. If you are realistic, you have to come up with
   something that works on top of existing operating systems (shudder),
   if you are revolutionary, you have to present a completely new
   concept, but you can hardly call it UN*X anymore, as most of
   what makes UN*X a fast and efficient system to work with is
   intimately related to its data structure: the ASCII text file,
   composed of English alphabetic characters.

						Thomas.