Quoth Dmitry Kurochkin on Jan 18 at 11:00 pm:
> On Wed, 18 Jan 2012 18:30:36 +0000, David Edmondson <dme@dme.org> wrote:
> > On Wed, 18 Jan 2012 22:04:36 +0400, Dmitry Kurochkin <dmitry.kurochkin@gmail.com> wrote:
> > > On Wed, 18 Jan 2012 17:39:31 +0000, David Edmondson <dme@dme.org> wrote:
> > > > The `mm-inlinable-p' and `mm-inlined-p' functions work better if they
> > > > have access to the data of the relevant part, so load that content
> > > > before calling either function.
> > > > 
> > > > This fixes the display of attached image/jpeg parts, for example.
> > > 
> > > Not so long ago I made an opposite change to avoid fetching useless
> > > parts (e.g. audio files).  Looks like we need a better check here.  Can
> > > we know from Content-Type if fetching a part body would be useful?
> > 
> > What if `notmuch-show-insert-part-*/*' consulted a list of content-type
> > regexps to attempt to inline?
> > 
> > That would allow a sane default (("image/*" "text/*") perhaps), but also
> > allow more to be added to that list (or some to be removed), either by
> > code that detected the (in)ability to render it or the user.
> 
> Perhaps there is such a list in mm already?

Shouldn't we only be doing this for parts with inline (or not
attachment) content-disposition?  That's cheap to check.  Or do we
actually want things like image attachments to get inlined, despite
their disposition?