• To: Thierry.Bouche@ujf-grenoble.fr
• Subject: Re: Comments on 0.56
• From: David Carlisle <david@dcarlisle.demon.co.uk>
• Date: Tue, 13 Jan 1998 18:46:15 +0000 (GMT)
• Cc: math-font-discuss@cogs.susx.ac.uk

```
> Now there is also a little package by D.~Carlisle that hides the dot
> with a white box, the problem being it is not background color
> independent...

But a reasonable number of documents are printed on white (or at least
constant) background, so this is more of a theoretical than a real
problem. The real problem is that the white box has a nasty habit of
obscuring surrounding text if you make it too large, and not obscuring
the complete dot if you make it too small. Mainly I did it that way so
it would work in non-PS drivers that support colour, eg Y&Y's dviwindo.

> Maybe the best device independance is achieved when you fake missing
> glyphs with graphics macros like \reflectbox? (if you clip or reflect
> with graphics, that could work at least in pdftex for instance,

but for dotless j you'd need clipping. I suspect (but don't know) that
it would not be too hard for driver writers to implement a clipping
rectangle \special, so that graphics.sty could supply \clipbox and
then dotlessj could easily be faked in a device independent
way. Currently I know of no dvi driver that supports clipping in a
native way, and so it would only be available in PS drivers (where the
literal postscript special can be used to add a clippath at the PS
level), so currently clipping of arbitrary TeX boxes is not provided
by the package.

Whatever `missing glyph' marker is used, my own preference would be
that the tfm metrics for that slot represent (an approximation to)
the metrics for the glyph that is missing, rather than the size of
the marker. This means that if someone wants to fake the glyph (eg
remove a dot from a j, or invert some symbol, or whatever) they at
least have reliable information about the size of the glyph that needs
to be faked.

David

```