Bugs:

* I haven't come up with a reliable way to distinguish command synopses 
  from stuff that should be treated as plain text.  This has a couple
  of consequences:

  1. doclifter doesn't recognize command synopses in displays, only in 
     Synopsis sections.

  2. When command synopses have a trailing text section (explanatory
     paragraph) doclifter doesn't always detect that and cope well.

* .Xo/.Xc Berkeley macro needs work. See elf.5.

* Look at lwres_getaddrinfo.3 -- good type case for resuming function parse
  after text.  sox(1) and soxmix(1) show the same problem with command syntax.

* CA.pl.1ssl -- CA.pl is not properly tagged as a command.

Edge cases in which we could do better --

* sox(1), soxmix(1)
    These are totally hosed.

Other things to do:

* Stereotyped .in/ta/ti pattern in emacs.1

* Stereotyped .in patterns under DIAGNOSTICS in compress.1.

Big projects:

* Translate EQN to MathXML.

See the BUGS file for other, more minor problems mainly due to bad markup.