Answers to common administrative questions about CL error handling

[Kent M Pitman <KMP@STONY-BROOK.SCRC.Symbolics.COM>]

I've been getting a lot of queries about the state of the condition system
lately and I wanted to get the community brought up to date...

 - Yes, version 18 of the condition system (AI.AI.MIT.EDU:COMMON;COND18 TEXT)
   is still `current' in the sense that there has been no version 19.
   Version 18 was approved by X3J13 and can only be modified by the cleanup
   process.  The only coherent conditions document which will follow will be
   the full ANSI spec.

 - Yes, AI.AI.MIT.EDU is hard to get to.  With some help from Larry Masinter,
   I have moved the two files holding the condition system to a new home
   on ARISIA.XEROX.COM, where you can get them via anonymous FTP...

     /cl/conditions/cond18.text
     /cl/conditions/cond18.lisp

 - Yes, there have been cleanups which have affected the correctness of
   version 18. The two most relevant cleanups are CONDTION-RESTARTS and
   CLOS-CONDITIONS, also available from ARISIA.XEROX.COM, in the files:

     /cl/cleanup/passed/condition-restarts
     /cl/cleanup/passed/clos-conditions

 - Given that version 18 has been amended and there is no coherent document
   to use, what should you take as an authority?

   If you are an implementor, you should definitely look at BOTH the
   version 18 proposal and all condition-related cleanups.  There may
   be others which have a smaller impact as well, but the two above are
   the most major.  Of course, how much you implement and how fast is
   really up to you--there is no ANSI spec yet, so strictly speaking
   there is not yet anything for you to say you conform to.

   If you're a user are are just doing programming in some system
   which claims to support the emerging system, my feeling is that
   you should look at NEITHER.  I recommend instead that you consult
   vendor-supplied documentation to find out what your vendor intended
   you to be using.  Although not really secret, these X3J13 documents
   are not public in the sense that they are not easily accessible.
   Also, they are prone to change.  As such, they are not something you
   should rely on as authoritative.  Different vendors are providing
   different levels of compatibility in the interim, and are tracking
   changes at different rates.  Only your vendor can tell you what
   you can reasonably expect out of their implementation.

I apologize for the broad distribution--I hope this information is helpful
to some of you who are scrambling to bring yourselves up to date. If you
have more questions, feel free to reply to me PRIVATELY. To avoid excess
mail traffic, do NOT direct your replies to the `X3J13' mailing list.
Thanks.