This file lists known bugs.  Or things you might think are bugs but are
actually supposed to be that way.  If you find one that isn't listed here,
please mail it to mcelrath+filterproxy@draal.physics.wisc.edu.

Netscape bugs:
--------------

* I have garbage text where a banner ad used to be!
  This is actually a bug in netscape not properly decompressing gzipped
  content, when that content is *inside* another page.  (using layer or
  iframe tags)  I have reported this to netscape (fat chance they'll fix
  it, or respond...).  Turning off Compression will work around it.  Since
  about 0.15, FilterProxy should filter out layer and iframe style ads.
  No other browser I know of has this problem.

* FilterProxy reports "Exiting for client ..." after receiving a 301 (Moved
  Permanently) or 302 (Moved Temporarily) response, or my browser acts like
  its still loading something (forever).
  Netscape forcibly closes the proxy connection after receiving these redirect
  responses, even though it is told not to by the 'Proxy-Connection: Keep-Alive'
  header.  It doesn't do this for 304 responses though.  Mozilla doesn't close
  the connection, but often (not 100% of the time) will just sit there waiting
  for more data.  Nothing in the HTTP spec indicates that the connection should
  be closed after these responses.

* FilterProxy reports "Exiting for client ..." after 1 connection on some pages.
  Netscape exhibits some bizarre behavoir on certain very complex pages.  In
  particular, it will load the entire page normally (without closing connections
  to FilterProxy), and then proceed to reload images in the page multiple times
  (up to 10 times PER IMAGE!), closing the connection after each reload.  Two
  pages which exhibit this are http://www.fileclicks.com and
  http://www.icewalkers.com.

* FilterProxy reports small images (~100 bytes or smaller) being reloaded in
  circles hundreds of times!  Netscape (and Mozilla) exhibit pathological
  behavior WRT small images (though I've also seen this bug with small HTML
  files too).  It will reload 1x1 gifs forever (example: greendot.gif on 
  slashdot.org), and will close the connection prematurely when loading these
  files.  This was claimed to be fixed in mozilla on 2/16/2001.  See bug:
    http://bugzilla.mozilla.org/show_bug.cgi?id=37589

These bugs in netscape have been present for a long time.  Netscape/AOL/Time
Warner have stopped development on Netscape 2 years ago, and concentrated on
Mozilla.  These bugs may or may not be present in Mozilla (I have personally
verified that several are absent).  I have made repeated attempts to notify
Netscape of these (and other) bugs, and I have never gotten a single response
out of them.  Please, I'm begging, SOMEONE write a better browser...I hate
netscrape, but I hate Microsoft worse...

It is possible to modify FilterProxy so that the "Exiting for client" bugs
don't have such a serious performance impact.  (by using the pre-fork server
model a la Apache)  This may happen in the future.  I am loathe to do a lot of
work to work around Netscrape's pathological behavior though.

FilterProxy bugs?
-----------------

* FilterProxy processes never exit!
  As of 0.27, FilterProxy switched to a different method of checking on its
  children (one that doesn't cause segfaults occasionally).  As a consequence,
  FilterProxy will leave zombie, or <defunct> processes.  These are "reaped"
  at the beginning of each request.  They are not a problem.

* [403 Forbidden] responses give me a [500 Broken pipe] when loaded.  If
  reloaded, they give the proper response.  (???)

* Many sites that attempt to do browser detection fail?  Why?  (even when no
  headers are filtered)

* HTTP Headers appear in response
  This is a bug.  If you see it and are able to reproduce it, please inform me.

* 407 Proxy Authorization gets associated with the site in mozilla.
  I have reported this as a mozilla bug.  See:
    http://bugzilla.mozilla.org/show_bug.cgi?id=69059

BAD URL's:
----------

http://www.webvan.com
    This page is entirely javascript.  The designers of this page should be shot.
    They use javascript to check which browser you're using, and then dynamically
    write the page for your specific browser.  Talk about unmaintainable.
    But seriously, I've added a Rewrite rule to 'ignore WEBBUGS' and 'ignore SCRIPTS'.
    The problem is that any rule with a -growto tag <script> will remove the entire
    page!  Any other website which (foolishly) chooses this mechanism of delivering
    web pages will have the same problem.  If you're a web developer, start writing
    for version 5/6 browsers now!  Throw off the Netscrape 4 cruft and hacks like this!
