Welcome to version 1.4 of NCSA HTTPD.

The *.tar.Z files contain the full distrubition.
The *.Z files contain only the httpd binary file.

The binaries are from the following systems:
httpd_1.4_aix3.2.5	RS6000/550 	running AIX 3.2.5
httpd_1.4_hpux9.0.5	HP 715/80 	running HP-UX 9.05
httpd_1.4_sunos4.1.3	Sun 690MP 	running SunOS 4.1.3
httpd_1.4_solaris2.4	Sparc 20  	running Solaris 2.4
httpd_1.4_osf3.0	Decstation 	running OSF/1 3.0
httpd_1.4_irix5.2	Indy	  	running IRIX 5.2
httpd_1.4_linux		Pentium90 	running Linux 1.2.5

We want to thank all of you who helped beta test NCSA HTTPD 1.4.
 - NCSA HTTPD Server Development Group
   Beth Frank, Stan Guillory, Brandon Long, Carlos Varela, and Yuxin Zhou.


Version 1.4 over 1.3R:
	- More security fixes
	- 50-100% performance increase
	- Webmonitor safe mail script
	- Configurable Error messages and scripts
	- TZ envirionment variable passed to scripts
	- redirect from scripts no longer takes 3 seconds
	- support for more systems, including QNX, SCO SVR3.2

Version 1.4 fixes from 1.4b6:
	- fixed test-cgi.tcl to make it CGI 1.1
	- added Webmonitor safe mail script
	- return default type on scripts when none is given
	- error_log now differentiates between SIGPIPE and SIGALRM
	- initialize_request now called for inetd and 1.3 fallback
	- rfc931 is fixed (its been broken all through 1.4 till now)
	- HPUX inststr should now work to change name of running process
	- Changed error script environment variables to REDIRECT_ header
	  to conform with Apache version of same

Version 1.4b6 does the following:
	- Lots of open file descriptor leaks fixed in nph,ssi,and other places
	- environment array memory leak 
	- TZ now passed in environment to scripts (by overwhelming request)
	- content-type now set to text/html if none given by script
	  (not sure if its the proper behaviour, but it seems to be necessary
	   for location redirects)
	- QNX and AUX support, improved NeXT support
	- HPUX signal handling for CGI
	- changed search pattern to better distribute load over children
	- fix to cfg_getline so it stops at eof, instead of reading 8k eofs

