sendpetpacket() needs to be able to handle larger than 256bytes packets.
	- ie fragmented transactions.

add man page

add example for swatch
add example for nocol
add example for sendmail with only aliases
add swatch to ftp site
add nocol to ftp site

setup all signals when starting
	- be paranoid about the state of the environment according to sendmail.

add timeout for sending mail

add timeout for pipe reads

handle SIGPIPE gracefully.

add maxtransactions option so can limit transactions to 1 or 2 per call
(which may be imposed by the provider, in which case sendpage needs to
fail once then try again)

if a multi transaction send fails, all the individual transactions should
be marked as single-only send -- the pet protocol has an abandon code
that it can, after the pages have been accepted decide to not accept
them, so when sending multiple pages a failure can happen such that
all pages are marked invalid, even though only one actually caused the
error.

modem init code doesn't work too well - sometimes takes 2 tries.

truncated pager files that queue == corrupt queue file

add a lastresort option.  Ie. if the page can't be sent for some reason,
forward it on to another machine.

add pc=dumb for old-style non pet/ixo/tap protocol paging
	- call the number send a few tones.  non reliable, etc.

disable id.pagingcentral by default for mail

only the sendpage user should be able to start the daemon.

add stats (smothers@mcd.intel.com)
	># of pages attempted
	># of pages sucessfully sent
	># of failed pages
	>    BUSY
	>    BAD PIN

use gnu autoconf

all the
	#ifdef DEBUG
		report
	#endif DEBUG should turn into

	RDEBUG(90, LOG_INFO, "Whatever message")

Look into modem/WaitString problems from keithr@aiinet.com
	Message-Id: <199509071215.IAA08210@locust.net.ohio-state.edu>

memory leak when expanding group aliases?

config parser doesn't always report the correct line on error

jcarolan@mayo.EDU reports QueueExpand() failure at about 40 messages
in the queue.  I still can't reproduce this problem without 
low disk space and/or low free inode conditions :(  My only guess so far
is that some implementations of mkstemp() may be broken.

add in includeAlias option to include another alias file

a zero size splitSize should equal no split

maxSplitMsgs should be configurable per user.

>   I put a complete tar file with the AIX changes and
>   options in /pub/sendpage on this system, if you're
>   interested.  A few minor modifications were needed
>   to get it to compile. - daryl@tcomeng.com

steve@demon.net:
> * Messages seem to go out of order.

"Jason A. Dour" <jadour01@homer.louisville.edu>:
* Ideas on message splitting.  Have you ever thought of sending '1/X',
  '2/X', et cetera, instead of just '1', '2', and so on?  That way, the
  recipient would have an idea of how many to look out for.  Also, should
  the msgIncludeSender be off for that user, perhaps a "tracking number" of
  some sort would be useful for those messages, so the recipient could
  thread through them should they receive other pages in between.

* Documentation.  I know you don't want to get a ton of documentation down
  yet since it's alpha...but I really can't decipher portions of the
  sendpage.cf.  How do you do user profiles?  Are they searched by alias
  name or by #######.pc1 or by what?  I'm really confused by the profiles.

steve@demon.net:
 * the email notification should report that the message was split
   (i.e. 1 of 2)
 * messages are split half way. Should try to do end of line, full stop
   or similar.

nabil@teleport.com  
	digital unix patches

sleep() in a few of the recent contributed patches have probably broken
the other timers.

The hangup function is getting stuck in a close() system call sometimes
and doesn't wake up until a USR1 signal is recieved from the sendpage client
on the next message.  This looks like another SunOS serial driver bug.

