[imp] IE "page cannot be displayed"

[Lord Apollyon]

> in your configuration files?  We've had this in our configuration files
> for years, and we're still getting sad email from IE users on both Mac and
> PC.

Hrm... c-client is usually pretty decent about handling screwed-up headers
and whatnot... it'll return "bogon" values like INVALID-USER at INVALID-DOMAIN,
etc.

Some email clients are even screwing up the RFC-mandated Date: header line,
the most common mangling I've seen is displaying the UTC offset without a
sign... i.e., 0600 or 0800.  *F***ED* 

Ah well.

In any event, I don't see "Page cannot be viewed" errors at all with mostly
IE5.x users and a small handful of IE6/XP users.  Do you have a session
cache configured for your SSL-Apache server?  IE seems to borf sometimes
without one.  There are three main things you need to deal with to handle IE
SSL brokenness:

1) Downgrade the response headers to HTTP/1.0 and prepare Apache to handle
the unclean shutdowns that IE does.  Go through your Apache/SSL config and
make sure the setenvif REGEX pattern is .*MSIE.* and not some
version-specific string... I've seen the SSL brokenness for certain in 5.x
and 4.x versions.

2) Disable the TLS EXPORT56 ciphers.  Note, do *NOT* disable ALL of the
TLS/SSLv3 ciphers, as some insane websites suggest... the only problem ones
are the lame SGC/56-bit ones... now that the US has let go of its insane
policy on limiting encryption, everyone's using 128-bit SSL capable browsers
now.

3) Configure a working SSL session cache (any driver method, MM, DB, etc). 
I have seen some weirdness with MM, so you may want to use a file-based
method, and put those SSL session files onto /dev/shm or similar "ram-based"
filesystem.  The locking is done correctly at least.

Make sure you have enough memory configured for your PHP process - 8MB + 2 x
the max attachment size is a good rule of thumb.

*THIS* is one area where I'm very unimpressed with PHP... it's a bloody pig,
and I am still confused as to how+why in the year 2003 that such things
still insist upon insisting upon loading the entire "attachment" into core
memory, then operating on it (in core memory), and then sending it to the
client.  My MAXMEM is up to 24MB now, just to handle my SMTP-configured
maximum of 8MB email sizes.

Ah well...
 
=Apollyon=