for both read and write.

Simplify the bouncing of e-mails using smb_updatemsg() and the original
message's from_ext convenience pointer.
No need to call smb_init_idx() here since smb_addmsghdr()->putmsg() does
this already (in the latest smblib).

-Deuce

"to" field as the new messages "to" field, by default.

now does this for us).
Set the SMB recipient extension header field when routing to QWKnet node
(this is necessary for fixsmb to be able to reconstruct the index file).
Use new function smb_updatemsg() to mark sent SMTP mail as deleted (much
easier).

dir->ex_ar is NULL, which shouldn't normally occur, but could in the BBS temp
dir menu.

(for time and message number) if the index hasn't been initialized.

no need to do manual initialization here.

that they match. This will catch a QWK networking bug in previous
Synchronet versions where the index.to was non-zero (the number of
the QWKnet node account), but there was no hdr.to_ext field, thus the index
could not be recreated from the information in the header (e.g. by fixsmb).
This was actually an intentional design decision at the time, but in hindsight,
it's a bug (the index file is not supposed to contain any unique information).

status display (percentage complete).

Wasn't accounting for correct lengths of keywords: WIN32, UNIX, LINUX, and PROT
Didn't support >8 char internal codes in SUBCODE and DIRCODE values
Would skip pass the line (null) terminator causing stack corruption or crash
when parsing values for SUBCODE, DIRCODE, or SHELL keywords as the last
keyword.

a 404 instead (ie: http://www.your.bbs/index.html/whee/this/is/cool)

variables since it MAY switch to CGI later.

CGIDirectory is specified.

Add REQUEST_URI CGI variable (Apache extension)
Drain STDOUT as well as STDERR
Don't drain stdio until script is terminated
Do not send Content-length header for CGI scripts.
Properly handle partial writes of CGI output using sock_sendbuf()
Should this be in sockwrap?

records (using smb_putmsg) - keeps the index file from being truncated during
the operation (nightly purging of deleted message indexes?) causing corruption.

msg.total_hfields to 0).

sent headers.

Don't directly access the index (.sid) file stream, reading a record at time
with no locking. Instead, read all the relevant index records in one safe
locked operating (using loadmail()) - this should fix the multi-server heavy
traffic mail.sid corruption bug.

in dumpindex().

of the host.  (ie: index.cgi)

This would allow index.cgi to be (say) viewcvs.cgi and URLs such as:
http://cvs.synchro.net/src/sbbs3/websrvr.c
To work.

to webctrl.ini

Now reads EVERY matching key with keys at the beginning of the file being
overriden by later keys.

[*.cgi]
Realm=CGI Scripts

[*.*]
ARS=LEVEL 90

Would have both applied to *.cgi and level 90 applied to everything
regardless of order in the .ini file whereas:

[*.cgi]
Realm=CGI Scripts
ARS=Level 50

[*.*]
ARS=Level90

Would behave differently than:

[*.*]
ARS=Level 90

[*.cgi]
Realm=CGI Scripts
ARS=Level 50

In the former, the ARS=Level 50 line would be overridden by the ARS=Level 90

on a per-dir or per-filespec basis.  Allowed keys are:
ErrorDirectory	- Changes the dir error pages are served from
CGIDirectory	- Changes the CGIDorectory value
ARS		- ARS required for this resource (Overrides access.ars)
Realm		- Authentication realm name to send to remote

Filespecs are the section name and the first match wins.  So:
[*.cgi]
Realm=CGI Scripts

[*.*]
ARS=Level 90

Would mean that for CGI scripts, the ARS would not apply.

And:
[*.*]
ARS Level 90

[*.cgi]
Realm=CGI Scripts

Would never set the Realm name.

section) or per service, using the LogLevel key (default value: LOG_DEBUG).
This commit requires an xpdev/ini_file.* update.