Re: libmisc and applications

["Michael K. Johnson" <johnsonm redhat com>]

Andrew Morgan writes:
>The problem with putting it in a public place is that applications are
>then going to become Linux specific (Sun is not offering this
>library), but since the stuff in misc is (hopefully) compatable with
>libpam of Sun and we are providing the source, it doesn't seem that
>bad.

Well, putting all the applications in apps/* doesn't make them any
less Linux-specific.  Even if libmisc *weren't* publically available,
applications pamified with it would be easier to port to the public
interface.  And since it will be publically available and should be
interoperable with sun's libpam, I think that you are right that it
doesn't seem that bad.  As long as we give it an appropriate name
to avoid collisions, I think we are set.  I think that we should use
"pam" in the name so that it is clear that this is a helper library
for pam.

>On a related issue (and one I've asked for before), if someone wants
>to make a misc_conv function that uses xlib calls[...]

I'd like to suggest that it would be better to use existing code
from each X app so as not to change its user interface significantly.
My plan is to accept username and password, store them in memory,
and then *before the window is unmapped* (so we can deal with errors
nicely in the same window) call pam to do the authentication,
with a conversation function that just returns the values already
typed in.  (I know that this doesn't map perfectly to PAM, but it
will work for the first pass.  How many X apps will be doing both
Kerberos and unix logins (one case where my idea wouldn't necessarily
work perfectly)?  None that I can think of.)  I'm doing this for
imapd, where the client sends a line like
a001 login username password
and then I have a conversation function that assumes that echo off
means password (it's really the only time it's called, because the
username gets passed in with the pam_start() call, but I've written
it to handle anything because I expect it to be useful with other
programs later) and returns the password when asked.  I intend to
put this into the X programs, at least for the first cut.

That's not to suggest that a good x-based misc conversation function
wouldn't be a good idea -- I just wanted to show what I'm going to
do in the meantime to deal with the problem in a way that will be
useful at least in the short term.

Good luck getting the new release out!  I'm looking forward to it.

michaelkjohnson

"Ever wonder why the SAME PEOPLE make up ALL the conspiracy theories?"