Mark Sapiro writes:
If mailman generates web pages with non-ascii, say utf-8 encoded characters and the installation's web server assigns a default character set other than utf-8, all the utf-8 encoded characters will be garbled. This will not happen if the characters are encoded as HTML entities.
So have Mailman do the encoding. Mandate that the input data such as templates are UTF-8, that template writers are responsible for ensuring that entities (eg, &) are syntactically correct (they will not be HTML-escaped), the output HTML pages are ASCII, and have Mailman do the translation of non-ASCII to HTML entities. Patrick gets what he wants, Mailman's generated pages are robust to webservers with broken encoding configs, and it is simple to check whether a user has done something stupid like encode their templates in Windows-1252.
Note that this can be done algorithmically by using the horribly ugly Unicode "entities", and the extension to prettier named entities is easy.
You could even provide a config option to turn off HTML escaping (defaulting to escaping on). If somebody has the skills to look at HTML source in a browser and cares, they probably have the skills to configure a webserver properly.
Steve