On Tuesday 15 June 2004 23:47, David Farber wrote:
> How sucessful do you feel your translation was?  How happy were you
> with the result?  'auto' does seem awkward at first, but it is not
> _that_ bad--it falls in with 'autoconfiança', 'autocontrole' and
> 'autodidata'.

I wrote a reasonably long Smalltalk application on paper using this 
translation (a bytecodeless Smalltalk-in-Smalltalk which was later 
converted to English here http://www.lsi.usp.br/~jecel/st84.txt) and 
was pleased with the result.

Luiz Fernando Bier Melgarejo came up with an interesting solution to 
this problem for the Agora language, implemented by Marcio Q. Marchini 
on top of LindaTalk (http://www.magma.ca/~mqm/mqm-cv.html). "Self" has 
three different roles: as a receiver for action messages, the receiver 
for accessor messages and as an argument. He did not attempt a single 
translation for all cased but had

  devo crescer   ([I] should grow)
  meu conteudo   (my content)
  minha cor      (my color)

I don't remember what "self" was translated to (if at all) when used as 
an argument.

> [Apple IIGS]

I was just kidding - see how much Squeak grew from 1.13 to 3.7, meaning 
what I did is such a small fraction of what would be needed today that 
it isn't worth trying to recover it.

> English also let you compound words more aggresively than the
> romanace languages.  So we get ReadWriteStream without much effort. 
> What did you come up with in Portuguese?

I just added "de" and "do" making it a little longer than the English 
version. "Stream" and "String" were particularly difficult to translate 
and I don't remember right now what I used for them. ReadWriteStream 
could be FluxoDeLeituraEEscrita, for example.

> So back to our original post--what would it take to translate
> Smalltalk/Etoys to Portuguese?  How hard would it be to run a simple
> word counter on the source code and spit out, say, the 100 most used
> words (after splitting keyword:messages: and ClassNames) and then the
> 100 most used method names.  That might be a place to start.  That
> might also be a nice tool to have in general for other translation
> efforts.

There are four levels in inscreasing effort:

1) eToys vocabularies - this has been included in the system for a few 
years, now (see methods in EToyVocabulary) and has Dutch, English, 
German, Kiswahili(?), Norwegian, Spanish and Swedish as of Squeak 3.6. 
You just have to copy a very long method and edit the two tables it 
includes in order to do a translation for a new language.

2) Babel - Diego's new system which requires small changes to the code 
which generates text for user interface elements. I think Squeak 3.6 
(but certainly 3.7) includes his patch so that Babel can also provide 
the translation for eToys when the vocabulary thing mentioned above 
doesn't exist for a given language. The actual tools to do the 
translation aren't part of the basic Squeak but are in a separate 
package you can download. You don't need the tools just to use a 
translated image. The idea is to eventually include every UI text in 
this system so that all menus, ballon help messages, buttons and so on 
can be translated.

3) many years ago I did a quick test in Smalltalk V of what I called 
MultiSymbols - this was a subclass of Symbol where more then one String 
could map to a single MultiSymbol. The reverse mapping (when pretty 
printing, for example) was controlled by a global. This was for 
translating Smalltalk itself, and allowed you to type in a German 
Smalltalk code and then read it in Spanish. Comments weren't translated 
(which, sadly, doesn't make as much of a difference as it should given 
how much Smalltalkers actually comment) and neither were UI texts (so 
something like Babel would still be needed). This would only solve the 
problem for the core Smalltalk - any new symbols created when 
programming applications would not normally include translations. The 
pretty printing system in Squeak is rather complex, so this would take 
a lot more work than it did in Smalltalk V.

4) a proper translation system - with everything stored in an 
intermediate language (like the UN universal network language 
http://www.ias.unu.edu/research/unl.cfm) and rendered automatically 
into the user's natural language on demand.

I feel that any translation must be complete even if within some limited 
domain (eToys only, for example). A partial translation is actually 
worse than nothing since whatever is left untranslated will be terribly 
frustrating for the users who wouldn't even have tried the system if it 
didn't seem to be available in their language.

-- Jecel