I agree that a) is probably best. See e-g/i-f-o just like a glyph in a word, for example. At least, it could be used that way. In most situations you'll probably surround the graphic with spaces anyway, so break possibilities come from there.
On 09.11.2005 05:51:03 Manuel Mall wrote: > What's the opinion around the group on how external-graphics / > instream-foreign-objects are suppose to be handled with respect to > determining linebreak opportunities. > > a) There is no intrinsic linebreak opportunity on either side of an > e-g/i-f-o (of course if surrounded by spaces or other breaking chars > these will produce a linebreak opportunity) > Knuth sequence: > box w="..." > b) They act more like a word surrounded by zero width spaces, ie one can > break before and after. > Knuth sequence: > pen w="0" p="0" > box w="..." > pen w="0" p="0" > c) Like b) but its undesirable so we penalise it, like a hyphen. > Knuth sequence: > pen w="0" p="FLAGGED" > box w="..." > pen w="0" p="FLAGGED" > d) Some (weird) combination like only allow a break after.... > > a) is certainly the simplest and an author can always put a ZWSP around > an e-g/i-f-o element. But would the "average" user expect it to behave > more like b) or c) (FWIW - MS Word behaves more like b)? On the other > hand for b) and c) we need an override if an author doesn't want a > break, ie. an explicit zero width joiner would be required. I am > tending towards a). > > Manuel Jeremias Maerki
