>Document where? If they are internal, it's for a reason, no? Don't >cast an API in granite if there is no need. But certainly the source >code should talk about what's going on.
I think the API has already been cast? Here's my thinking. We support postproc and have for a while, yes? To write your own postproc you need to know what the various undocumented switches are, so your postproc can at least know how to deal with them (you need to handle some of them special, and at a minimum you need to know which ones take an argument). This means to write an effective postproc you need to read the source code to understand what is going on. This is a pain in the ass. I've done that already, and it makes me wonder ... why is this crap not written down? I mean, if there is an interface why do you have to trawl through the source code to use it effectively? I'd hate to have to make someone else waste a day trying to understand what -idanno does. And maybe someone will think of something cool to do with it that I never thought of. Looking at the complete list of undocumented switches, I really didn't understand the reasons for them all being that way. Like, -server was undocumented for a long time; why? And we've changed the API, actually; a bunch of switches were garbage collected a long time ago. To me this is more about documenting how the world works. --Ken _______________________________________________ Nmh-workers mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/nmh-workers
