Subject: Re: Barriers to Lisp acceptance - a "survey" question
From: Erik Naggum <erik@naggum.no>
Date: 1999/03/02
Newsgroups: comp.lang.lisp
Message-ID: <3129400264626689@naggum.no>

* Kent M Pitman <pitman@world.std.com>
| (Of course, you probably find out by latency or net protection errors,
| but no equivalent system not based on CORBA does any better.)

  dealing with networking problems is so complex that an application really
  cannot afford to trust a pre-packaged solution that does not give very
  good access to the error conditions.  I've spent hundreds of hours giving
  Allegro CL's socket package and my protocol on top of it graceful failure
  modes.  it's really hard work.  judging from what I see from people who
  just use some random pre-packaged library, this is not an area that
  people tend to worry about, hence we get very dangerous solutions that
  work in a perfect world, but not in the hostile world we actually live
  in.  what little I know about CORBA has not shown me a condition system
  that is capable of what I need from a basically unreliable universe where
  your application no longers dies when the machine dies, it may get wedged
  in some zombie state when some machine you didn't know about dies, like a
  router, a firewall, or simply someone messing up routing tables, etc.
  stuff like happens all the more often the more you are interconnected,
  and a distributed object system is no good if it doesn't deal with such
  problems gracefully, in my opinion.
  
#:Erik