J.Henshall@EDINBURGH.AC.UK (06/19/89)
I am a little unhappy at the points Christian raises in support of X over ACSE. Listed as the first advantage is that it conforms to the architecture! Even as a firm OSIist I find it difficult to see that as an advantage in its own right. Certainly to place X over Transport sort of 'breaks' the model but, if the reasons are good, then that is fine by me; we are trying to find solutions for the real world. The second advantage is that it would work over 'single package' implementations like IBM's without modification. Are we really going to be making design decisions based upon conformance with proprietary implementation techniques? Surely large multi-purpose stacks such as IBM's are intended to provide Session Services to all manner of applications, many of which will be driven in background (spooled) enviroments where performance and responsiveness are less critical. As such then these stacks may well be wholly inappropriately implemented to form the basis of a 'real time' style of application like X. The third advantage concerns packing of many X requests into a single "S (or T)" SDU. This speaks for itself...it can be done with T. Steve Kille raises the points of authentication and encryption. Here we have possible candidates for use of the high layers. Don't get me wrong, I am not trying to set the ISO architecture up as a target for shotgun practice; I just think that sooner or later we will decide to 'permit' the Transport interface to open up for direct service anyway, and will then regret pushing things up to the top out of reverence for the model. John (Am just off for a couple of weeks holiday in Germany so will not be around to field flak for a while)