Karl Berry wrote:
> Therefore, when you deal with an URI, you should use a different
> algorithm of presentation within a GNU error message than when you
> deal with a filename.
>
> This seems like it adds complexity both in the description and in
> comprehension.
The long description of the algorithms was only meant to prove that
someone has gone through all the details.
> At least I am having trouble with it -- it seems like %
> now means two different things,
No, not at all. % means one thing: The proposed use for filenames and the
use of % in URIs and URLs are the same. The difference is that when you
pass a filename, it is not yet escaped. Whereas an URI is always transferred
in escaped form between programs and machines.
> and a program has to know in advance
> whether the source is a url or a filename.
Yes. This is a consequence of Henri's requirement, who wants that URIs
and URLs be copy-pasteable and human-readable.
> As an alternative, we could simply use a different character than % for
> our filename hex escape.
This is not simpler:
- If the other character you use for escaping is one that can occur in
URIs, such as '!', it means the presentation contains two layers of
escaping, the one with % from RFC 2396, and the one with ! on top of it.
Less simple, and does not guarantee that URIs and URLs are copy-pasteable.
- If the other character you use for escaping is one that can not occur
in URIs, namely '<' or '>' or '"', then URIs and URLs are copy-pasteable,
but the perception for the user is that filenames and URLs are escaped
in two different ways: filenames with '"' as escape character, and
URLs with '%' as escape character.
In summary, my proposal is easier to understand for the end user (since it
uses a single escape mechanism for filenames and URLs), while the programmer
has a little bit more work: he has to pass a boolean argument to specify
whether he uses a filename or an URL.
As always, I value simplicity for the end user higher than simplicity for the
programmer.
Bruno
