[Dancer-users] [ proposal ] charset discussion

[damien krotkine]

Hi,

This email is to share some thoughts I had while reading  this commit :

https://github.com/sukria/Dancer/commit/6e30f2fd23a9ab00cb45d54e13556ee30b841f99

I was willing to fix a minor thing (the fact that the feature provided in
the commit works only with utf8, not any other encoding), then I realize
that there are 2 concepts of charsets that are mixed in Dancer (at least I
think) :

- the 'charset' setting is used to set the response headers' encoding and
charset (see
https://github.com/sukria/Dancer/blob/master/lib/Dancer/Handler.pm#L129 )
- the 'charset' setting is used to read files (with this new open_file
feature)

These 2 concepts are legitimate, but having them bound to the same setting
is error prone : let say you have latin1 template files, and you want your
Dancer application to provide UTF8 encoding, you can't. And I can see people
trying to fix that by mixing the charset setting, and the ENCODING options
of D::P::TemplateToolkit for instance. That will be messy.

The Dancer documentation says that the 'charset' setting is "The default
charset of outgoing content". I think that it should be kept this way. the
'charset' setting should not be used for local file reading/writing.

Also, Dancer::Renderer::templates provides a default layout template (
https://github.com/sukria/Dancer/blob/master/lib/Dancer/Renderer.pm#L200 ),
but set its charset as UTF-8 by default, which contradicts the 'charset'
documentation : "Default value is empty which means don't do anything. HTTP
responses without charset will be interpreted as ISO-8859-1 by most
clients.". But maybe I missed something, especially the cases
where Dancer::Renderer::templates is used.

Anyway, I can see the current situation bite us in the back. Unless I have
misunderstood the code, in this case, let me know :)

These steps could fix the situation :

- open_file() should not refer to the same thing used by the Handler to set
Dancer headers
- there should be a way to indicate to Dancer that local files read /
written are in a given charset
- also, my personal opinion is that the 'charset' setting could be utf8 by
default, as utf8 is cool :) But that's not mandatory to fix the issues

Sadly, I don't know how to do that without adding a new setting, something
like 'local_charset', or 'files_charset'... which would be used y open_file,
the 'charset' be left as it is now, and used only for Dancer responses.

What do you think? I have a branch halfway done, but I'd like to see more
inputs and opinions, so that we achieve a common and agreed strategy :)

dams
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.backup-manager.org/pipermail/dancer-users/attachments/20101214/0faded0d/attachment.htm>