Filled in a function comment.

[privoxy.git] / doc / changes.txt

This file contains details of the changes made to JunkBuster, in
chronological order.  Scroll down to the bottom for the newest 
additions!


*****************************************************************************
* Copied from old README                                                    *
*****************************************************************************

README for the Internet Junkbuster Proxy (TM) Copyright 1997-8 Junkbusters Corp.

   Id: README,v 1.1 2001/04/16 21:10:38 rodney Exp 

   This program is free software; you can redistribute it and/or modify it
   under the terms of the GNU General Public License as published by the
   Free Software Foundation; either version 2 of the License,
   or (at your option) any later version.

   This program is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
   See the GNU General Public License for more details.

   You should have received a copy of the GNU General Public License
   along with this program; if not, write to the Free Software
   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.

The version of the code with this makefile is Version 2.0.2.
No CHANGES file is included here; for a history of previous versions see
   http://www.junkbusters.com/ht/en/ijbdist.html#previous
This version tightens security over 2.0.1; some multi-user sites will need
to change the listen-address in the configuration file (see previous URL).

For installation on Windows see http://www.junkbusters.com/ht/en/ijbwin.html
An executable is provided so you don't need nmake or a C compiler.

For UNIX and other systems see the FAQ. If you just want to test it with its
defaults of simply stopping cookies and private info, all you need to do now
on most Unix systems is type make, run junkbuster & and configure your browser.

This directory should contain the following files.  The .html files are also
in http://www.junkbusters.com/ht/en/ and http://internet.junkbuster.com
You are welcome to publish copies of the .html files on your local web server,
according to the GNU General Public License.

gpl.html GNU General Public Licence
junkbstr.exe   Executable (binary) for Windows 95/NT
Makefile A very bland Makefile; Most versions of Unix except HP & Suns
      compile without change. Windows requires the changes indicated.
ijbman.html Manual for the Internet Junkbuster, HTML format
junkbuster.1   Manual for the Internet Junkbuster, man macro format:
         man ./junkbuster.1 # or nroff -man junkbuster.1 | more
ijbfaq.html FAQ: Frequently Asked Questions on the Internet Junkbuster
      Includes installation instructions
junkbstr.ini   Sample configuration file with almost all options commented out
sblock.ini  Sample blockfile (which doesn't really block much, see FAQ)
scookie.ini Sample cookiefile (permits little; you decide who to trust)
saclfile.ini   Sample access control file (block access to everyone)
sforward.ini   Sample forwardfile (does no fowarding, no gateways, so SOCKS)
strust.ini  Sample trustfile (has no trusted sites and blocks nothing)
conn.c, bind.c, encode.c, loaders.c, parsers.c, ssplit.c   Various utilities
socks4.c Code for SOCKS4 and SOCKS4A gateways
jcc.c    The main program
jcc.h    #included declarations etc
gnu_regex.c, gnu_regex.h:  Regular expression code from the FSF, not us.
win32.c     A few lines of code specific to the Windows version (95 & NT)
acl.c    Access controls (new in version 2.0) - doesn't replace firewalls
filters.c   Strings for the presentation of messages to user

If you find that the Internet Junkbuster improves the quality of your life
online, we hope you'll visit http://www.junkbusters.com to see how our free
services can help you bust other kinds of junk out of your life:
junk mail, spam, telemarketing, and invasions of your private data.
And please tell your friends about this free product, and help them set it up.

If your company finds this software useful in protecting its confidential
information and making its people happier and more productive, please consider
purchasing one of our our commercial support packages:
   http://www.junkbusters.com/ht/en/ijbfaq.html#commercial

We don't advertise, so please tell others in your community about us.

Junkbusters Corp.    http://www.junkbusters.com


*****************************************************************************
* Copied from old README.too                                                *
*****************************************************************************

Id: README.TOO,v 1.1 2001/04/16 21:10:38 rodney Exp

This is the README.TOO file for Stefan's version of the Internet
Junkbuster (from http://www.waldherr.org/junkbuster/).



Installation:
-------------
Install the RPM with the usual command `rpm -Uvh foobar.rpm'. Since we
run the Junkbuster as user `nobody', cd to a directory where `nobody'
has read access (in particular NOT /root) and issue (as root)

   /etc/rc.d/init.d/junkbuster start

This will be done automagically for you any time you boot your machine.



Configuration:
--------------
To let Junkbuster take `care' of the ads, you have to setup your
browser to use the Junkbuster.

  Short instructions for the advanced user:

   Configure Netscape to use 127.0.0.1 on port 8000 as 
        a proxy server.

  Detailed instructions for the beginner:

        Remove Internet Explorer and install Netscape. No `real' 
        user would ever use MSIE. Though the Junkbuster works 
        with IE as well. Read 
        http://www.junkbusters.com/ht/en/ijbfaq.html#browser

Check to see if the Junkbuster works correctly by surfing to

   http://127.0.0.1/show-proxy-args

or any other place and add the text `/show-proxy-args' to the URL.
Ok? Then enjoy your ad-free surfing.



Mailinglists:
-------------
In case you are interested in the Junkbuster, there are two
mailinglists (junkbuster-users and junkbuster-announce). Instructions
on how to subscribe and unsubscribe are at

   http://www.waldherr.org/junkbuster/

Bugs:
-----
Please fill out

   http://sourceforge.net/bugs/?func=addbug&group_id=11118

Patches:
--------
Please fill out

   http://sourceforge.net/patch/?func=addpatch&group_id=11118

Need Help?
----------
Please fill out

   http://sourceforge.net/support/?func=addsupport&group_id=11118


Comments:
---------
I would greatly appreciate feedback for this program. Send comments to
Stefan Waldherr <stefan@waldherr.org> (in either German or English). I
prefer plain text. If you happen to see an ad which got thru the filter
PLEASE use the `contribute' form at http://www.waldherr.org/junkbuster/.

Have fun,
Stefan.


*****************************************************************************
* Copied from old README.win                                                *
*****************************************************************************

$Id: README.WIN,v 1.1 2001/04/16 21:10:38 rodney Exp $

This is the README.WIN file for the Windows version of the Internet
Junkbuster (from http://www.waldherr.org/junkbuster/).

Installation:
-------------
Congrats, since you are reading this file, you have already unzipped
the archive correctly.

Start the Junkbuster with `junkbstr' (i.e., click on the icon in the
ijb subdirectory). If you want to start Junkbuster automatically,
make a link to the program in the `Autostart' folder.



Configuration:
--------------
To let Junkbuster take `care' of the ads, you have to setup your
browser to use the Junkbuster.

  Short instructions for the advanced user:

   Configure Netscape to use 127.0.0.1 on port 8000 as 
        a proxy server.

  Detailed instructions for the beginner:

        Remove Internet Explorer and install Netscape. No `real' 
        user would ever use MSIE. Though the Junkbuster works 
        with IE as well. Read 
        http://www.junkbusters.com/ht/en/ijbfaq.html#browser

Check to see if the Junkbuster works correctly by surfing to

   http://127.0.0.1/show-proxy-args

or any other place and add the text `/show-proxy-args' to the URL.
Ok? Then enjoy your ad-free surfing.



Mailinglists:
-------------
In case you are interested in the Junkbuster, there are two
mailinglists (junkbuster-users and junkbuster-announce). Instructions
on how to subscribe and unsubscribe are at

   http://www.waldherr.org/junkbuster/

Bugs:
-----
Please fill out

   http://sourceforge.net/bugs/?func=addbug&group_id=11118

Patches:
--------
Please fill out

   http://sourceforge.net/patch/?func=addpatch&group_id=11118

Need Help?
----------
Please fill out

   http://sourceforge.net/support/?func=addsupport&group_id=11118

Comments:
---------
I would greatly appreciate feedback for this program. Send comments to
Stefan Waldherr <stefan@waldherr.org> (in either German or English). I
prefer plain text. If you happen to see an ad which got thru the filter
PLEASE use the `contribute' form at http://www.waldherr.org/junkbuster/.

Have fun,
Stefan.


*****************************************************************************
* Copied from old README.re_filter                                          *
*****************************************************************************

DISCLAIMER:
   This is pre-alpha code. Using it in any other way
   than just reading it is a high risk activity and
   the author disclaims all liability for any damage.

   See the GNU General Public License at
   http://www.gnu.org/copyleft/gpl.html for details.


Dear alpha-Tester,

Thank you for trying out the experimental re_filter feature of
Junkbuster. As this is still a very early pre-release version,
and I'm neither much of an autoconf or make wizard, building
the patched junkbuster involves some manual work, that is 
described below. Please also see the Request for Thoughts section,
as the whole point of this pre-release ist to collect ideas for
improvement and bug reports.


INSTALLATION:

  - Get and install pcre from
    ftp://ftp.csx.cam.ac.uk/pub/software/programming/pcre/
    Version 3.1 or any later one will do. I recommend 3.4.
    It *should* install on Win32 as well, but I haven't tried
    that. Don't forget to run ldconfig afterwards.

  - Now you have two options:

    A) Install pcrs as a shared library:
       pcrs.c and pcrs.h are included in the (patched) ijb subdirectory.
       1.) patch Makefile < Makefile.re_shared.diff  
       2.) gcc -Wall -fPIC -c pcrs.c  
       3.) gcc -shared -Wl,-soname,libpcrs.so.1 -o libpcrs.so.1.0.1 pcrs.o -lc -lpcre
       If this worked OK,
       4.) copy libpcrs.so.1.0.1 to /usr/local/lib
       5.) ln -s /usr/local/lib/libpcrs.1.0.1 /usr/local/lib/libpcrs.so
       6.) ldconfig (Or whatever the equivalent on your platform might be)
       7.) cp pcrs.h /usr/local/include

    OR

    B) Incorporate pcrs into the junkbuster (much easier):
       Smile and relax. You're done.

  - Edit the provided re_filterfile to suit your needs and place it
    with the rest of junkbusters configfiles.

  - Now junkbuster should build with a simple 'make'.
    Note: Setting debug = REF in jcc.c will enable debugging
    for the re_filter.

  - Include a line "re_filterfile <wherever>" in your junkbuster
    config file, pointing at the re_filterfile

  - (Optional:) Include a line "re_filter_all" in the config
    file. This will set the Policy to filter content from all sites.
    Default is to filter only from sites that you wouldn't send a
    cookie either. 

  - Fire up junkbuster.

HOW IT WORKS:

    Nah, not tonight anymore ;-)

REQUEST FOR THOUGHTS:

  - The main decision is whether to filter on-the-fly or to buffer
    up the whole document and then run the filter on it. I've built
    both versions and currently prefer buffering.

    The advantage of on-the-fly-filtering is obviously that the
    document begins to display before it's fully loaded. The big
    disadvantage is that with arbitraty content altering in place,
    you don't know the size that the document will have after
    altering when the Content-Length Header is to be sent. Only
    option is to chew it altogether.

    Apart from preserving the ability to generate a valid Content-Length
    header (which the patch currently does not yet do), buffering up
    the document has some advantages from the filtering point of view:

     - Reliability. Patterns in the document cannot be ripped apart
       and are reliably found. (If you look at the document fragment-wise
       while it's being transmitted, you miss any pattern that spans
       a fragment border) This is especially bad for patterns that do
       serious HTML reformatting, as fragments are typically in the order
       of 1500 bytes each.

     - Flexibility: The re_filter gives you the choice whether you want
       to replace only the pattern's first occurance or replace it globally.
       This becomes meningless with fragments. (Solution possible, though.)

    As the filter only processes objects of MIME type text/*, which are typically
    rather small compared to binary multimedia (sic) type data, I don't think
    that the disadvantage of delayed display of the first bytes is really grave.

    What do you think?

  - Is the choice between a radical (filter all) and semi-smart (filter only when
    also crunching cookies) flexible enough? Should I really introduce yet another
    configfile for specifying the domains to filter content from? Would anyone use
    that?

  - Should I use pcre as a shared library and work out a nice installation
    automagic, or should the relevant functions be incorporated in re_filter.c?

  - Even if yes, do we really want an extension that relies on an external library
    like pcre?

  - Any bugs, stupid parts in the code ?

  - Cool suggestions for the re_filterfile? (PLEASE!)

  - Should character set translation along the lines of Perl's 'tr' operator
    be included?

  - Is the whole feature just pathetic?

FEEDBACK:

    A whole lot of work has gone into this feature (the pcrs code was also
    just written for this) and I'd really like to know if and how you like
    it. Since I'm more of a perl than C programmer (and after doing so much
    string manipulation in C I'm once again glad abaout that!), I'm sure there
    is much to improve. Please send any suggestions/observations/hellos to
    oes@paradis.rhein.de. I would also be very glad about any assistance for
    a autoconf/make procedure for prcs.

Well, have fun!
--Andreas


*****************************************************************************
* Copied from old README.cygwin                                             *
*****************************************************************************

Id: README.cygwin,v 1.2 2001/04/29 23:16:29 rodney Exp


What I did:  I saw too many "personalities" in IJB; so I
ran the code through a couple of regexp's and then let XEmacs
reformat the whole shebang!  The results is a uniform
looking code base; aka. as if 1 person created the whole
project.

Why I did it:  (1) it is easier to maintain  (2) easier for
a new person to make changes/additions  (3) it *looks* more
professional  (4) IMHO, it is easier to read (this is very
much a personal preference).

What I propose: A IJB standards list be included in the
distro and all source/patch checkins be checked for
compliance.  We don't need to be an arse on this point, but
submitters should make a valiant attempt to support the
standards.

Note: do not even try to diff my code with another code
base, you will be overwhelmed.  After all, the regexps did a
good bit of changing and XEmacs reformatter did a great deal
more.

One more note: these changes include conditional compilation for
un?x (namely Linux), cygwin (un?x like compile for WIN32), and
mingw32 (WIN32 like compilation w/o VC++).  Right now, all seems
fully functional except cygwin.  The cygwin compile will work,
but includes code to make it single threaded.  I need to debug
the fork issues, but that will have to come later.


Specific changes to expect:


1: I've majorly reworked the Makefiles.  The existing Makefile was
too *spaghetti* for me!  For one, I eliminated the "-I." from the
include path (ugly).  Use #include "" for local directory files
and #include <> for path found files (that's basic even to c).


2: In accordance with #1, I changed #include "windows.h" to
#include <windows.h> and #include <gnu_regex.h> to #include
"gnu_regex.h".


3: As part of my standards suggestions (to come later in this
file), I added a standard comment block to every function.  There
were too many functions with *no* comment and about 20 different
styles to those that did.  This is an example of the block I used:
/*********************************************************************
 *
 * Function    :  block_acl
 *
 * Description :  Block this request?
 *                Decide yes or no based on ACL file.
 *
 * Parameters  :
 *          1  :  src = Address the browser/user agent is requesting.
 *          2  :  dst = The proxy or gateway address this is going to.
 *          3  :  csp = Current client state (buffers, headers, etc...)
 *
 * Returns     : 0 = FALSE (don't block) and 1 = TRUE (do block)
 *
 *********************************************************************/


4: Some files prototyped like this:
      void
      func(param1)
      int param1;
      {
      }
and other like this:
      void func(int param1)
      {
      }
I made all look like the second.


5: Made:
      if(  to be  if (
      while(  to be  while (
      switch(  to be switch (
      etc ...
The former looks too much like a function call.


6: Put braces on their own line and aligned under the
reserved word:
      if () {
      } else {
      }
to be:
      if ()
      {
      }
      else
      {
      }
The same for while, do, etc...


7: Casted malloc in several places, e.g.:
      p = malloc(n);
to be:
      p = (char *)malloc(n);
This eliminates compiler warnings.


8: Applied the Perl regexp patch and made it fully conditionally
compiled (-DPCRS=TRUE).


9: It *appears* that if MINGW32 sees a "main" function, it will
try to link it and start with main (ala console mode).  Since I
need MINGW32 to find WinMain instead, I conditionally changed
"main" to "_main" (for MINGW32 only!).


10: Changed popup.c to be killpopup.c.  This was necessary to
clean up the Makefile.  Since there was an ini type file called
"popup" and "popup.c", make got confused on target names.  No
other ini file has a .c file with the same name, so I hope this
isn't a big deal.  Plus, it fits its config variable
(kill_all_popups) much better.


11: Added:
  /*
    Local Variables:
    tab-width: 3
    end:
  */
to all .c and .h files.  I use 3 character tabs and this will make
the files display correctly in Emacs editors.  This is just my
default setting.  Perhaps I will filter all files via "expand
-t3", so not to pi$$ everyone off.  Vi is capable of using 3
character tabs, as I am sure VC++ would be.


12: Several places included regexp code with out "#ifdef REGEX"
which broke the ability to compile w/o regexp's.  I fixed the code
and can compile w/o regexp (which I don't normally do, but now I
*can*).


13: In WinMain, changed:
      pszLastTok = strchr(pszLastTok, ' ');
to be:
      pszLastTok = strchr(pszLastTok+1, ' ');
The former causes an infinite loop with 2 or more parameters.


14: Added my own strdup for Cygwin, see jcc.c for why.


15: Added defines in the Makefile and this to jcc.h:
      #define VERSION VERSION_MAJOR "." VERSION_MINOR "." VERSION_POINT
This makes "update the numbers also in jcc.h manually" unnecessary
(see your Makefile for this quote).


16: Split the "inifiles" target in Makefile into separate targets,
this way each file can be regenerated individually according to
the same implicit rule.


17: With the Makefile restructure, I can only provide for
Cygwin/gcc.  Hopefully this is a bit cleaner and others can add
their compilers w/o re-spaghetting the file.  I have provided
several makefiles as examples only, I do not intend for the
project to start maintaining multiple Makefiles.  Perhaps some of
the platform specific Makefile code could be split out into
separate files and we could use the "include" directive to read in
the correct one.

Better than this, let's get a configure.in guru on the project and
dynamically create the Makefiles!


18: In an effort to get this to market (you knew this was coming
eh?), I am going to put the standards suggestions in another file
and ship this one as is.  But then, you can probably guess most of
the standards by the changes you see.


19: Added a really quick and dirty "install" target to the
Makefile.  This needs to be cleaned up and proper directories put
in for each platform.


-- release 1 (obsoleted) --


20: Incorporated the TOGGLE and webDAV support patches into my
Cygwin patch.  I added conditional compilation to both features.
I changed "extern BOOL g_bToggleIJB;" to be of type "int".  This
makes jcc.h independent of "windows.h".  The inclusion of such
should not (and is not) necessary for compilation on all
platforms.  Besides, BOOL is really only a char or an int anyway!


21: Added a new ini!  I modified the usual un?x type config file
to M$ type ini file for a Perl regexp compile.  This was necessary
to call EditFile on the Perl regexp config file.  The new ini
filename is "sregexp.ini" (naturally).  I then added menu items to
edit the popup.ini and sregexp.ini files.


22: I made "dot h" files for all the "dot c" files.  I used this
command to do the *dirty* work:
    grep -e "^[A-z]" *.c | sed -e "/=/d" -e "/^[^(]*$/d" \
         -e "/^extern/d" -e "/^static/d" -e "/;$/d"

And I made each "dot h" file be immune to multiple inclusion via
conditional compilation.  Ie. the "jcc.h" file will only include
itself if _JCC_H is undefined.

After this was done, I compiled the project source by source and
let the compiler tell me what each file was dependent on.  Once I
got all the objects to compile, I included the new dependencies in
the Makefiles.

BTW, I intend to get the Makefile to be generated by a configure
script.  Whether this is a GNU "configure.in" type file or a shell
script, I do not know.  All I do know is that this is a must for a
"Go/No go" type decision.  We cannot proceed with the philosophy
"leave it up to the user to get it to work".


23: I broke out the code to load the config file from main.  I put
this code into a function called load_config.  I made this routine
be called from a signal "HUP".  This means we can restart
junkbusters from the command line or from the daemon script via
"restart" without actually killing it and restarting it.  This is
common behavior to un?x daemons (see inetd for example).  I fixed
the junkbusters.init script to take advantage of this.  I intend
to make a M$ menu complement to this as soon as I can.


24: BTW, if you have not already noticed by my postings and/or
code, I have added patches and made them all conditionally
compiled.  Ie. #ifdef TOGGLE, #ifdef WEBDAV, #ifdef PCRS, etc...
I think that all extensions to IJB should be submitted in this
way.  Not only can you compile IJB to be as "tight" as you want,
it also gives the user a "smorgasbord" of options to
choose from.  This is quite common in the free software world (you
need look no farther than emacs to see this) and provides maximum
flexibility.


25: As promised, I obsoleted the "ijbw32.ini" file.  I was a bit
confused by the "close-behaviour=1" entry.  It seems to me that it
duplicated the "close-button-minimizes=1" entry.  And (by looking
at the code) it was unused.  Anyway, I eliminated the "ijbw32.ini"
file and assimilated the settings into the config file.  Bye Bye
"ijbw32.ini", hello "config" (aka. don't unnecessarily duplicate
files and/or data types).


26: Eliminated re_filter.c.  This only defined
(un)load_re_filterfile and re_process_buffer.  I moved
(un)load_re_filterfile to loaders.c and re_process_buffer to
filters.c. I also modified the Makefiles to reflect this.  In the
doing, I modified the loader and unloader to use the "files"
variable and moved joblist to "re_filterfile_spec" struct.  I
added "plist" to the client_state so that all Perl regexp code
could find the latest parsing of the re_filterfile ini.

This is in keeping with the IJB data structures philosophies.


27: Eliminated acl.c.  This (as pcre) only defined
(un)load_aclfile and block_acl.  I moved (un)load_aclfile to
loaders.c and block_acl to filters.c.  I also modified the
Makefiles to reflect this.


28: Since I created "dot h" files for all the "dot c" files, I
eliminated all the "extern ...variable;" and all the "extern
...function;" statements that I could find in the "dot c" files.
Instead, I included the proper "dot h" files to do this for us.


29: Corrected the "install" Makefile target for un?x.  I will do
the same for WIN32 targets.


30: Note on the "TOGGLE" patch: since I made IJB fully restartable
via a NOHUP signal; I thought the TOGGLE patch should apply to
both un?x and WIN32.  Thus, I eliminated the WIN32 dependencies in
that patch and added "toggle" to the config file.  Now everything
works fine for either platform.  Still to come is a restart menu
command in WIN32 (See: note 23).


31: Just a possible programmer note, we REALLY need a generic
linked list module.  Since text lists, client states, filter
files, etc... implement list creation and destruction over and
over; I suggest a generic list that can handle *any* type of list
(ala template classes in C++).  If this module included an
iterator (of sorts), we could eliminate the myriad of "for ( top=x;
NULL != x->next; x=x->next )" type statements that are all over
IJB.

After all, this was taught in the basic "Intro to Data Types"
class (for all you CS majors).


32: Renamed all of the .INI files to be .TXT files.  I did this
because the .INI files were NOT *really* ini files; which
potentially be confusing.  I used the .TXT extension to make sure
that EditFile still worked.


33: Removed the "initialed comments", such as "/* swa */".  This
leads to less readable code (because it becomes fragmented) and
retains the "multiple personality" syndrome.  Comments such as
this can always be figured out from the versioning system.  Also,
these comments really should be pooled in one place : the README
files.


34: Perhaps we should break with the current versioning tactic of 
a "tail version" and start fresh.  Perhaps a new major version
release is in order?  Version = 3.0.0?


35: Added a "Accept-Encoding: gzip" cruncher.  If a stream is
compressed via gzip (Netscape specific I think), then it cannot
be modified with Perl regexps.  So I added this as an option in
the Makefiles.


36: As promised in one of my postings, I have make a
configure-esk script to generate the Makefiles.  I would still
like to have a full blown configure.in compatible system, but this 
will do in the meantime.  Note: this configure script covers
linux, cygwin, and mingw32.  At this point I do not have VC++ to
test any of my changes.  But the configure script should be
flexible enough to handle this platform.


--

Please give any feedback to the ijb egroups and, if you feel you
need quicker contact, CC me at IwantToKeepAnon@yahoo.com

Let me say once more, this has a *lot* of changes and may (aka
WILL) break a lot of existing patches.  But I hope the "idea" will
catch on.  Perhaps this could be made the baseline code base and
other patches added and standardized later.

*****************************************************************************
* Date:        7 May 2001                                                   *
* Version:     ??                                                           *
* Description: Changes for MS Visual C++                                    *
* Author:      Jon Foster <jon@jon-foster.co.uk>                            *
*****************************************************************************
(Based on Stromlund's version 0.20)

I have made several changes to make this compile under VC++ 97, and add new
features.

Code changes:

1) If you disable IJB (using the TOGGLE patch) it now turns PCRS off.
2) "Missing function prototype" warnings have been corrected by adding
   appropriate #include statements
3) The _DIST_URL define removed and hardcoded.  There are hardcoded URLs all 
   over the place that need fixing - one more doesn't matter.  We do really 
   need to come up with a long-term solution though.  (Suggestion: Host a 
   redirector on the SourceForge web space.  e.g 
   "http://ijbswa.sourceforge.net/redirect.php?ver=2.0.2-10&target=faq",
   "http://ijbswa.sourceforge.net/redirect.php?ver=2.0.2-10&target=download",
   etc.  This allows us to move the pages around easily, and the SourceForge
   space can be maintained by anyone who's interested, it's not tied to a 
   particular individual.  This is similar to the approach taken by 
   Microsoft and Netscape in their browsers.)
4) New file ijbconfig.h is #included everywhere.  This provides me with a
   place to define VERSION, PCRS, etc.  You can still define VERSION_xxx
   on the command line, but the other defines are in the header file.
   There were just too many #defines - so many that the VC++ 97 IDE 
   broke when I tried to define them on the command line.  We probably 
   should fix the configure script so that it modifies this file.
5) Deleted empty file afxres.h - we needed to use the VC++ system header 
   file of that name.  However, it's easier to just delete it and put 
   the relevent #defines in w32res.h.
6) New #define: NO_PROGRAM_NAME_DISPLAY uses "IJB" as the program name in 
   the log file instead of "H:\Documents and Settings\Administrator\My 
   Documents\Prog\VC\ijb2\Release\junkbuter.exe" or similar.  The reason 
   for this should be obvious!
7) New #define: STATISTICS enables the statistics feature
8) Statistics is thread-safe, I think.  However, loading up the
   statistics page now counts as a blocked request.
9) New, shorter message if a setting in the config file is recognised 
   but unsupported by the current build.  It doesn't have "WARNING:" in
   front of it either.  We may want to actually take this message out 
   and have it just silently ignore these entries.
10) New feature: #define SPLIT_PROXY_ARGS and the show-proxy-settings 
    page will split each file out to a seperate page.  This is 
    because the files are quite big, and having to do 1 extra click to 
    get to them is no problem.  This #define also disables the 
    suppress_blocklists feature (because I believe the original intent 
    of suppress_blocklists was similar?).  It also saves memory by 
    reading the config files to generate the HTML on the fly, rather 
    than storing prepared HTML constantly in memory.
11) Marked global constants with "const" attribute.  This has a 
    knock-on effect on pointers in function prototypes, so these have
    also been marked const where needed.
12) Rewrote encode.c to have tables pre-initialized, so it can be 
    marked constant.  Also split into 3 distinct functions so that
    the tables (which are realy an implementation detail) are not
    visible outside that file.
13) Changes to how files are reloaded when you reload the config 
    file.  Previous method didn't free memory (or if you uncommented 
    the free() calls, it was not thread-safe.)  It is now handled 
    automatically - files are reloaded whenever the filename to 
    read changes, or the file's timestamp changes.  The code in
    loaders.c which handles this test has been moved to a single
    utility function.
14) killpopup changes:
     * Only compiled if you #define KILLPOPUPS
     * Call to filter_popups() moved from read_socket() to chat().
     * Writing 1 past end of buffer bug fixed.
     * It was scanning the whole buffer even if only partially filled.
     * Moved load/unload to loaders.c and changed file handling to 
       follow standard.
15) Auto-detect whether we want an image or HTML implemented for 
    Microsoft IE.  :-)
    There doesn't seem to be a way to tell with the latest build 
    of Mozilla. :-(
16) Image file code conditionally compiled.
17) Provided option (SPLIT_PROXY_ARGS) to split the show-proxy-args
    page into a main page, and then a page for each config file.  
    This option will also save memory by loading the files when 
    required for display, rather than saving them in memory when
    they are initially loaded.
18) Force page loading ("noijb.") patch applied.  #define FORCE_LOAD


Packaging changes:

1) PCRE 3.4 (less it's test and demo programs) is now included in the 
   pcre directory.  It will be statically linked with Junkbuster.  (On 
   win32, DLLs cause a lot of hassle - with a library this small it isn't 
   worth it.  I don't know if it would be better to dynamically link under
   UNIX?)
2) junkbustr.dsp and junkbustr.dsw VC++97 project and workspace files are
   included.
3) win32build/junkbustr.exe Win32 binary is included.  It should not require
   any DLLs.  It was compiled with all features supported.
4) All README.xyz files put together in this file.
5) Unused files (acl.c, pthread.c, ...) deleted.
6) My block & cookie files used (based on the Waldherr ones).

Known bug: The Adobe Acrobat 4 plug-in fails when I try to view PDF files in
MSIE.  Workaround: disable "web browser integration" in the Adobe Acrobat 
settings.

*****************************************************************************
* Date:        14 May 2001 (early morning)                                  *
* Version:     Reported as 2.9.0                                            *
* Description: Various updates                                              *
* Author:      Jon Foster <jon@jon-foster.co.uk>                            *
*****************************************************************************

1)  Now use PCRE, not GNU REGEX.  I have not yet had chance to check the
    syntax of the block/image/cookie file to ensure that they match what
    is expected - however they seem to work.
2)  Replaced "configure" script with one generated by "autoconf".  Also 
    use a header "config.h" (was ijbconfig.h in my previous release) for 
    the #defines.  "config.h" is now generated with "autoheader" from 
    "acconfig.h" and "configure.in".  (Note that to install you do not
    need autoconf or autoheader - just run "./configure".)
    To see command-line options, run "./configure --help".
    This is my first ever autoconf script, so it has some rough edges
    (how PCRE is handled is the roughest).
3)  Error logging code replaced with new module errlog.c, based on the
    one from JunkBusterMT (but with the threading code removed).
4)  Most of Rodney's 0.21 and 0.21A patches applied. (Marked *).  I did not
    apply all of these, since I had already independently done conditional
    popup file, conditional image file, and integration of popup code.
5*) ACL, Jar and trust files conditionally compiled.
6*) New source file headers.
7*) Various cosmetic changes.  (But I have not consistently ordered the 
    config files - I think that's worthwhile, but it's 1am and I want to
    get this released!)
8*) RCS tags on .h files.
9)  RCS tags are const char[] rather than const char *.  (Saves 4 bytes
    per tag ;-)
10) VC++ project files renamed to vc_junkbuster.*.
11) show-proxy-args now shows status of all conditionals, not just REGEX
12) Various functions moved around.  Most notably all the system-specific
    sockets code which was spread between jcc.c, bind.c, and connect.c,
    has been moved to "jbsockets.c".  The non-system-specific code from
    connect.c and socks4.c has been movet to "gateway.c".  Also, the
    config file loader and the global variables it writes to have been
    moved to "loadcfg.c".  (Maybe this should go into loaders.c?)
    And candidate for the "worst filename ever" award is "miscutil.c",
    which contains, well, miscellaneous utility functions like zalloc.
    (Suggestions for a better name for this file are welcome!)
13) Loaders now use a common function to read a line and skip comments,
    and this function also stores the proxy_args.
14) Added ./junkbuster --help     (Not for Win32 GUI)
15) Added ./junkbuster --version  (Not for Win32 GUI)
16) Win32 resources are now all marked as "U.S. English", rather than
    being a mix of "U.S. English", "U.K. English" and "Irish English".
17) Version number changes to 2.9.0

Known bugs:
- See (1) above about blockfiles.
- Sending SIGHUP (under Red Hat Linux 7.1) causes a crash.  v0.21A hangs
  after a SIGHUP.  This needs investigating, but pthreads support will 
  almost certainly help here, so it may not be worth investigating
  immediately.
- Compiling with shared system pcre and pcreposix libraries is supported
  via a switch to configure, but completely untested.

Please note that there are now 2^14 == 16384 combinations of conditional 
defines, and 4 major compilers (VC, Linux, mingw32, cygwin) for a total 
of 65536 different builds.  We cannot possibly test all of them!  My 
standard build is "everything on", and I've tested that on each compiler,
along with a few variants.  You may uncover new bugs (probably compiler 
errors) if you use an exotic combination of switches.

*****************************************************************************
* Date:        14 May 2001                                                  *
* Version:     2.9.1                                                        *
* Description: Various updates                                              *
* Author:      Andreas S. Oesterhelt <oes@oesterhelt.org>                   *
*****************************************************************************
(From his e-mail:)

 - in parsers.c, fixed two #ifdef FORCE to #ifdef FORCE_LOAD
   (BTW: I think FORCE is precise enough, since loading remote
   data is the whole purpose of a proxy..)
 - Set the FORCE_PREFIX (back) to 'IJB-FORCE-LOAD-'. While 'noijb.'
   is more elegant and looks like a hostname in the URL, it doesn't
   make clear to the inexperienced user that the proxy is bypassed. It
   also has a higher name collision risk.
 - Filled in the function header templates for my functions in
   parsers.c (again). They obviously got lost in our current
   patch war ;-)
 - Cut the credit for the §-referrer-option from the config file,
   that Stefan had placed there.
 - Improved the re_filterfile 

*****************************************************************************
* Date:        14 May 2001                                                  *
* Version:     2.9.2                                                        *
* Description: FORCE patch (again!)                                         *
* Author:      Andreas S. Oesterhelt <oes@oesterhelt.org>                   *
*****************************************************************************
(These notes written by Jon)

- Andreas applied the latest version of the FORCE patch.

*****************************************************************************
* Date:        14 May 2001 (late afternoon)                                 *
* Version:     2.9.3                                                        *
* Description: Various updates                                              *
* Author:      Jon Foster <jon@jon-foster.co.uk>                            *
*****************************************************************************

1) Incorporated updates from current CVS tree, including:
   - Amiga support (completely untested by me - I don't have an Amiga)
   - "tinygif 3" support (redirects blocked images to a specified URL, so
     the browser doesn't have to load and cache many copies of the same
     image).
   - one case where there were both local and global "referrer" variables
     (yuck!) clarified by renaming the local one to "refer".
   - Fixed some places where close() was used instead of close_socket().
   Thanks to Jörg Strohmayer (joergs at users.sourceforge.net) for these.
2) Temporary hack to get FORCE_LOAD to work with IE.  I just lowercased the
   FORCE_LOAD_PREFIX.  Needs fixing properly.
3) Most URLs hardcoded into JunkBuster were changed to go through a script
   e.g. http://ijbswa.sourceforge.net/redirect.php?v=2.9.3&to=faq
   The only other URLs left are the GNU GPL:
     http://www.fsf.org/copyleft/gpl.html
   and the home page:
     http://ijbswa.sourceforge.net/
   ... and various URLs which will be intercepted by JunkBuster anyway.
   TODO: Still need to do something with the URLs in JunkBuster Corp's 
   copyright/trademark notice on the bottom of the show-proxy-args page.
4) PCRE or GNU Regex is now a #define option.

*****************************************************************************
* End of file                                                               *
*****************************************************************************