alpha 1 released jan 9, 1995
--
increased internal dialer default timeout from 20 to 25 tries when waiting
for answer

allow for early go ahead message when talking PET.  This is not part of
my protocol spec....

fixed hanging bug in config file parser.  Now also does "" processing so
fields can have spaces in them, which can also be done with \ 

sendpage -bp works of no pidfile present.

calculated and displayed delay of page received to page sent for
email response.

fixed missing close() in queue run.  This caused a fd leak and sendpage
would crash after running a few days.  Also caused the SunOS serial driver
to get stuck in a weird mode (sendpage would get stuck in an Exiting state)
requiring a reboot.

added a less strict pet lookup table conditional compile.

added uucp style locking, although there is no standard way to do this --
 some programe stick the pid in as binary, others ascii.  sendpage uses
 ascii, and if the pid doesn't exist, ignores the lock.

added PARANOID_REPLIES option to thwart possible attempts of exploting
sendmail's over-trusting nature.

When the administrative limit on pages per message has been hit, no
longer ignore entire message, instead just inform user that the rest have
been dropped.

incorrect installation will more likely exit with a sendmail exit
status that informs the user, instead of just the postmaster.


incorporated netbsd patches from Marc_Unangst@DENEB.FAC.CS.CMU.EDU
	termio.h -> termios.h
	bugfix in config.c to (sscanf -> fscanf)
	open modem device with O_NDELAY
	ioctl -> tcgetattr/tcsetattr
	missing comment delimiter in io.c
	SYSV_MALLOC is required to have mallinfo()
	fixed writev() call in queue.c

started merging in BSDI patches from Ben Golding <bgg@connect.com.au>
	added a conditional for malloc.h
	termios vs. ioctl() (done previously)
	checks modem return code for 2400bps (also added others)
	
started adding support for external dialer script
	(suggested by sslok@srv.PacBell.COM)

added code to hangup modem if paging central doesn't (untested)

added a little bit more documentation

added run time configurable per paging central message lifetime
	- ie quit at 60 retries

added run time configurable per paging central message size limit
	- ie truncate messages at 80 chars.


merged in matches from  Ben Golding for PG3 support (untested)
	protocol=pet-pg3


dtlist (delivery type list) was redundant and confusing - removed.

report() defaults to LOG_MAIL now, and works properly when in
daemon mode.

fixed potential bug with file aliases (chance of non null terminated
	pathname)

files that have no data result in the default alias being used now.

added pipe style aliases (ie - alias operator |/etc/calendar)

added sendpage_env variable and use execle() to give children a
known safe environment.

fixed malloc() size error in AddMessage() - would cause strange
	things if more than 1 or 2 messages were sent depending on
	how much space malloc() wasted at the end of a hunk.

---
alpha2 released Mar14, 1995
---

fixed m (max messages) option
	- should not have set STATUS_EXPANDED bit.

fixed sending pages to non alias destinations, ie 12345.usamobile2
	- recipient2len not set by QueueExpand.

fixed error reporting for STATUS_EXPIRED

added RETRIES_NOTIFY_FAIL config option.  Instead of always sending
a reply on the first failure, can now be configured.

---
alpha3 released Mar16, 1955
---

workaround for syslog() and report() static buffers.
	- report() uses a 4K buffer (same as syslog)
	- all calls to report with unknown size strings use %.512s instead of %s

config.c: pcinfo.modeminit not terminated properly
	- daryl@tcomeng.com

signal.c: added POSIX_SIGNALS code.  Use sigaction() for signal() instead
	of signal().

merged most of the AIX patches from daryl@tcomeng.com & mark@tcomeng.com

sendpage.h can control a little more of the modem hardware setup --
	specifically the DTR toggle for hangup and RTS/CTS flow control.

merged in HP/UX patches from Mike Lutz & Heath Kehoe @ norand.com

minor changes to compile under IRIX -- untested.

added Solaris compatibility -- untested.

fixed race condition in queue.c which would result in a spurious
warning message about not being able to unlink() a queue file.

---
alpha4 released May 23, 1995
---

recipient is included in e-mail reply if VERBOSE_MAIL_REPLY is off again.

If there were multiple entries for a pc queue'd and being processed and
the initial modem connect failed, and the next modem connect (or upto
the # of messages being processed) succeeded, then the messages after
the succeeded call would end up as
	sendpage[11658]: write(): Bad file number
...and nothing more :(

in deliver.c, this format spec will result in a more standard time format...
fprintf(FP, "QueueTime:   %d/%d/%d %d:%02d:%02d\n",  (.....)
instead of the %d:%d:%d you had.
    - Aaron Nabil <nabil@i.net>
(done)

in queue.c AddMessageQueue() was sending error messages to stdout instead
of using report()

Merged in OSF/1 patches + the extra debugging code.
	- haymanR@icefog.uacn.alaska.edu

Merged in Linux portablity -- read() can return EAGAIN which is probably
true for more than Linux
	- Jay Allen jay@fork.com

added a few references to other pager related sites and documentation

---
alpha5 released ??? 
---

