Hi, Lars.  Here is a suggestion to get rid of those spurious white line=
s that
are often tied up with the usage of MML parts.  In fact, the idea is no=
t new.

If someone writes:

        [text]

        <!#part ...>
        [inside]
        <!/#part>

as readability invites to, there is one end of line between [text] and =
the
MML start tag already, but surely another after the `>' and the [inside=
].
Another new line is also implied after the MML end tag.  That might exp=
lain
those three white lines between parts we too often observe! :-)

The idea would to specify that an MML start part tag may only be recogn=
ized
when alone on its line, that is, the `<' should not preceded by non-whi=
te
characters, and the '>' should not be followed by non-white characters.
Preceding whitespace could be used to indent MML directives at will acc=
ording
to the part nesting (and Gnus should generate the edit buffer with inde=
nted
MML), it should be ignored when reading MML back in.  Following whitesp=
ace
would be ignored while reading MML back in.  Of course, between `<' and
'>', white space and new lines could still be used at will for readabil=
ity.
The main point is that the end of line after the MML directive should be
swallowed and ignored.

Right now, I'm being forced into writing things like:

        [text]
        <!#part ...>
        [inside]

which is too cluttered to my taste, because I would like to see a white=
 line
while writing MML, where there would be a white line in the rendered me=
ssage.
Another current possibility is writing:

        [text]

        <!#part ...>[inside]

which is not very nice when [inside] is many lines, as the first of the
lines is forced into some awkward indentation in the editing buffer,
making it unnatural.  I have experienced this many times by now.  If the
end of line following the MML tag was plainly swallowed as part of the =
tag,
it would allow more nicer writing of MML, in my opinion.

A debatable choice to make is to swallow, or not, the end of line which
precedes the tag.  My opinion is that it has to be, if we want MML to a=
lways
stand alone on its line and be nicely indentable.  That would make two =
white
lines in the editing buffer when one is meant in the rendering, which is
not very elegant.  Or else, if that preceding end of line is not swallo=
wed,
then it might be best then to forget the idea of indenting MML, and just
consider whitespace which precedes `<' as fully significant, like it is=
 now.

But swallowing the end of line following '>' does not look that debatab=
le
to me, as I see it as necessary for editing to be nice, while avoiding
spurious white lines at later rendering time.

Once consequence of my suggestion is that <!#part> embedded in a paragr=
aph,
not being recognizable as MML, might not require `!' quoting.  It would
then be a bit easier to speak about MML in running text.  The only case
requiring quoting would be while giving whole line as examples.  Once a=
gain,
this choice is somewhat tied to swallowing preceding end of lines or no=
t.

--=20
Fran=E7ois Pinard   http://www.iro.umontreal.ca/~pinard