lei convert: remove redundant input_net_cb

[public-inbox.git] / Documentation / lei-security.pod

diff --git a/Documentation/lei-security.pod b/Documentation/lei-security.pod
index 4b712c2d4233ed652494b629e045702f8d87146c..8cbd8993456893b776087c7f68087f38c4243961 100644 (file)
--- a/Documentation/lei-security.pod
+++ b/Documentation/lei-security.pod
@@ -18,6 +18,9 @@ permissions support.
 It does not use POSIX ACLs, extended attributes, nor any other
 security-related functions which require non-standard Perl modules.
 
+There is preliminary support for "virtual users", but it is
+incomplete and undocumented.
+
 =head1 INTERNAL FILES
 
 lei runs with a umask of 077 to prevent other users on the
@@ -93,12 +96,20 @@ lei uses L<git-credential(1)> to prompt users for IMAP and NNTP
 usernames and passwords.  These passwords are not encrypted in
 memory and get transferred across processes via anonymous UNIX
 sockets and pipes.  They may be exposed via syscall tracing
-tools (e.g. L<strace(1)>).
+tools (e.g. L<strace(1)>), kernel and hardware bugs/attacks.
 
 While credentials are not written to the filesystem by default,
 it is possible for them to end up on disk if processes are
 swapped out.  Use of an encrypted swap partition is recommended.
 
+=head1 AUTHENTICATION METHODS
+
+LOGIN (username + password) is known to work over IMAP(S),
+as does AUTH=ANONYMOUS (which is used by L<public-inbox-imapd(1)>
+as part of our test suite).  AUTHINFO may work for NNTP, but
+is untested.  Testers will be needed for other authentication
+methods.
+
 =head1 DENIAL-OF-SERVICE VECTORS
 
 lei uses the same MIME parsing library as L<public-inbox-mda(1)>