still codings problem (saving) with babel.

[Stephen J. Turnbull]

Aidan Kehoe writes:

 > This is what happens when Latin-1 is decoded as UTF-8, since I changed the
 > UTF-8 coding systems to avoid transforming every octet that is not valid
 > UTF-8 to tilde. Basically, babel (or more probably, the URL-retrieval code)
 > isn’t looking at the MIME information about what coding system the data is
 > in at all.

IIRC url.el (I think that's what babel uses) returns a binary blob
plus header information, and leaves parsing the header and converting
the content to other code (such as w3.el).  So I would guess this is
babel (or some middleware like w3) making assumptions.

In general, I would not expect url-retrieval code to handle the
rendering, AFAIK the MIME content-type could be multipart.  It could
special case text/*, but that would make me feel very sick. ;-)  And
there might be no useful out-of-band information at all (ie, the
served header says "application/octet-stream").