Subject: Re: corba or sockets? From: Erik Naggum <erik@naggum.net> Date: 01 Nov 2000 14:47:36 +0000 Newsgroups: comp.lang.lisp Message-ID: <3182078856687989@naggum.net> * Wade Humeniuk <humeniuw@cadvision.com> | ISORMOSI (first time I've seen this acronym!) for me still has a | place. I have used it to describe a protocol tha was very messy and that definition was the basis for a clean reimplementation, so I won't knock the model too hard, but it doesn't work for the Internet, and one must be acutely aware of when to break that model. Sometimes you need a lift in that 7-story high-rise. | I will usually try to design a protocol which follows a layered | model, because I have to apply some technique to get where I am | going. Protocol layering is as hard as doing class inheritance right. The interesting thing is that you can start with the lowest layer and have a slot called "payload" that a subclass would interpret with sub-slots, or you could start with the highest layer and add slots in subclasses as you move down the stack. As long as the needs of a class are dictated from the outside, this is pretty easy. It gets real difficult to find the One True Layering when you move your own functionality around as you experience changing needs. | This is one of the weird parts I find about life is that I seem to | need flawed (religious? dogmatic?) views of the world to approach | the truth. I guess its called learning. Is there a way out of that | morass? Apart from dying? I don't think so. :) | As for the triumph of TCP/IP over ISORMOSI. I think it was things like: | | Unix | FTP being simpler than FTAM | SNMP being simpler than CMIP | HTTP | the Session and Presentation Layers in OSI | TCP/IP was mostly American | random blind chance | and should I say "Simpler is Better" | | that killed ISORMOSI (may it rest in layers). | | BTW, isn't ISORMOSI still kicking in Europe? Nah, even the OSI Profiles (like GOSIP) have moved to TCP/IP, but there are still some large commercial X.400 software vendors and service providers. Some believe that EDI needs X.400 to work. I used to study X.400. I made a guess that I would spend 10 years writing a fully compliant mail system based on X.400. Then I wrote a fully compliant SMTP-based mail system in three weeks, added MIME stuff experimentally (while I was still contributing to that work) and figured that if you had to spend more than a man-year on a mail system, you'd need to make it a highly successful mass product, and that was very unlikely to happen to anything as long as decent and simple mail systems were available essentially for free. I don't think I was too far from how people with real money, resources, and vested interests were thinking. Today, we have the Evil Behemoth doing about 10% of what X.400 offered, and they still don't comply with the necessary RFCs, so chances are they can't even _read_ the X.400 specification. The WWW idea hit the world with unprecedented force. It's a crying shame that HTTP and HTML had such staggeringly idiotic designs, and still do. If I were Tim Berners-Lee, I'd blame someone else for it, so I guess he couldn't find anyone who would accept that. But I digress. #:Erik -- Does anyone remember where I parked Air Force One? -- George W. Bush