Known bugs and restrictions in version 0.19
===========================================

IP over ATM
-----------

 - first ping freezes some systems (???)
 - kernel crashes badly if (remote) ARP server is set to local machine
 - clients don't always refresh their entry at ARP server (fixed ?)
 - IP interfaces must not be reconfigured while up
 - requires local ATM address to be set on interface 0 on startup (fixed ?)
 - encapsulation is broken (but works good enough for now)
 - atmarpd may not recover properly from link failures (but a resolution 
   request or such usually cures that) (fixed ?)
 - atmarpd shouldn't invalidate entries on the first timeout in valid state
   but rather try one revalidation request first
 - Arequipa leaves stale SVCs if atmsigd has disappeared
 - creation of the hard_header isn't "clean"

Signaling
---------

 - never initiates a RESTART procedure
 - doesn't implement T322 (i.e. STATUS ENQUIRY is not retried)
 - atmsigd sometimes terminates ungracefully when encountering errors in
   signaling messages
 - no support for repeated information elements
 - there's currently no way to obtain SVC traffic parameters indicated by
   the remote party
 - SSCOP doesn't update the window properly

ENI-155p driver
---------------

 - TX freezes on some systems when starting atmsigd and ilmid (???)
 - driver may hang under heavy load with LANE
 - shouldn't depend on sizeof(unsigned long) == 4
 - sometimes doesn't unqueue RX PDUs when being tortured with aping
   (or, more generally, seems to mis-calculate buffer usage after CR errors,
   so that close hangs)
 - SUNI loopback is untested
 - support for multiple adapters per machine is untested
 - doesn't handle unsolicited OAM cells gracefully
 - adapter may deadlock if buffer full of loss indications
 - doesn't properly zero pad area of AAL5 PDUs
 - driver should use bottom-half interrupt handler

ZN1221 driver
-------------

 - shouldn't depend on sizeof(unsigned long) == 4
 - still hangs from time to time when sending
 - support for multiple adapters per machine is untested
 - SONET/SDH framing support is untested
 - doesn't support single-copy in this release

ATM over TCP
------------

 - is reported to hang under load
 - signaling is reported not to work over ATMTCP
 - driver is very limited in functionality
 - system may crash if atmtcp is killed
 - can't run IP over ATM using ATM over TCP
 - ATMTCP interfaces can't be taken down

ATM sockets and "common" layer
------------------------------

 - single copy currently uses massive amounts of brute force to flush TLBs ...
 - SVC close may block forever (waiting for atmsigd)
 - polled devices are broken as designed (and may eventually go away ...)
 - some system calls don't return full SVC SAP information
 - poor integration with other network code

Miscellaneous
-------------

 - there are lots of known bugs not listed in this file (look for @@@ in
   the source)
 - there are *still* no "nice" applications
