Re: [Squeak-VMdev] Proposal: VM string encoding

[Bert F. <be...@im...>]

Am 02.10.2004 um 02:47 schrieb Yoshiki Ohshima:

>   Bert,
>
>>>   We'll need a way to use possibly different encodings for:
>>>
>>>   * File name
>>>   * clipboard
>>>   * keyboard input
>>>   and possibly
>>>   * default system encoding
>>>
>>>   Assigning 1006 through 1008 would be good.
>>
>> Can you describe the circumstances when these encodings are different
>> from the default encoding?
>
>   It is rather typical in this part of the world on Unices.  Older
> systems tend to use EUC and newer ones usually trys to use Unicode,
> but often those are mixed in one environment and possibly all
> combinations can exist.

Okay. But how should the VM know which one is actually in use? I don't 
know any system call to get a character set.

>   I don't know too much about Mac, but as far
> as I know, it has different set of file access APIs and none of them
> is good enough to be used exclusively.  That is why (as far as I
> heard) the Japanese Squeak running on Mac OS X cannot use the Japanese
> image file name because at first the VM uses the MacRoman version of
> file name.

Hmm, can't comment on that ... John?

>> Are they at least constant over a session?
>
>   Yes.  We can assume this.

okay ...

>> Or can the encoding differ for distinct clipboard accesses or filename
>> uses? If the latter, then a VM attribute would not be optimal.
>
>   For file name, it has to be set in the earlier stage of VM
> invocation, but the clipboard and keyboard input can be set later.
> So, potentially read-write VM attributes would do it.

... so why would we need to *set* it if it's constant? Can't the 
conversion happen on the image side?

- Bert -