Re: [Gnu-arch-users] Encoding handling proposal

[Marcus Sundman]

On Monday 30 August 2004 00:18, Alexey N. Solofnenko wrote:
> There can be a much simpler approach:
> A) We still need a metedata (and, if it is mutable, it should be
> versioned).
> B) Text files can have optional encoding attribute (or during first
> import the attribute can be guessed by the system).
> C) All client computers have files exactly as in the repository.
> D) "smart" patch/diff/merge use the encoding attribute to correctly
> calculate differences between two files (of the same encoding!).

This is more or less how it would work with "Auto-Filter" set to "false" in 
my proposal. (Except that in my proposal the encoding info was still 
mandatory instead of optional.)

> Without changing content encoding on the fly it works without any
> problems. Maybe it is a good idea not to allow content encoding
> changing, because a file with new content encoding is essentially a new
> file, which can still look similar to the previous one.

I would argue that changing the content encoding makes a file a "new file" 
if and only if the encoding is part of the semantics of said file.


- Marcus Sundman