On Thu, 26 Feb 2004, Alan Altmark wrote: > I'm confused by the RPC+LE remark. What problems have you encountered? > We've written many test programs in C on CMS using the SMAPI RPC calls.
What David said. But also ... Dr. Boyes presents a good summary. Let me add two important points about RPC things. Not assigning any blame, I find them to be heavy and fragile. I don't see anything intrinsically wrong with the remote procedure call concept, but in contemporary programming there are few things which have this combination of bulk and tendency to fail. (In Unix land, NFS and NIS are the two things which will lock up a host faster than anything else.) :-( Browser based Java comes close. [ducking from flames] But more fundamentally, SMAPI is ... well, it's an API. There is an overwhelming de-facto framework in the industry, "Just define an API and code to that.". But that's short sighted. Think protocols; think pipelines; think outside the API box. When it comes to interfaces, Application Program Interfaces ain't the whole story, and aren't always the best way to talk. But we do have an API, and that's great. We just need more. As David mentioned, a reference C program (with source!) would go a long way towards adoption of SMAPI. And speaking of such handy things as REXX/Sockets, a sockets interface to SMAPI along with the API would help a lot and would let it come out of that API closet. Please! Consider this all to be positive criticism intended to foster wider use of our faVorite platforM. -- R;
