On Fri, 13 Jul 2012, Jameson Graef Rollins <jrollins@finestructure.net> wrote: > On Sat, Jul 07 2012, Mark Walters <markwalters1009@gmail.com> wrote: >> +typedef enum { >> + NOTMUCH_SHOW_OUTPUT_HDR_DATE = 0x1, >> + NOTMUCH_SHOW_OUTPUT_HDR_SUBJECT = 0x2, >> + NOTMUCH_SHOW_OUTPUT_HDR_FROM = 0x4, >> + NOTMUCH_SHOW_OUTPUT_HDR_TO = 0x8, >> + NOTMUCH_SHOW_OUTPUT_HDR_CC = 0x10, >> + NOTMUCH_SHOW_OUTPUT_HDR_REPLY_TO = 0x20, >> + NOTMUCH_SHOW_OUTPUT_HDR_IN_REPLY_TO = 0x40, >> + NOTMUCH_SHOW_OUTPUT_HDR_REFERENCES = 0x80, >> + NOTMUCH_SHOW_OUTPUT_DEFAULT = >> + NOTMUCH_SHOW_OUTPUT_HDR_SUBJECT | >> + NOTMUCH_SHOW_OUTPUT_HDR_FROM | >> + NOTMUCH_SHOW_OUTPUT_HDR_TO | >> + NOTMUCH_SHOW_OUTPUT_HDR_CC | >> + NOTMUCH_SHOW_OUTPUT_HDR_DATE >> +} notmuch_show_output_t; > > Is there a reason we need to limit this to some pre-defined subset of > headers? Wouldn't it be nice if you could just specify arbitrary > headers? I basically agree but it looked like it would be relatively ugly to pass around. However, perhaps this is all being too general: the caller probably cares that all the headers it wants are output and perhaps not too many others (eg over ssh or to android etc). Would something more like headers=brief or headers=full be enough? And we would still want an option allowing the body to be omitted. What do you think? Best wishes Mark