> I'm confused by the RPC+LE remark.  What problems have you
> encountered?

It's a two part problem -- neither is directly the fault of SMAPI, but both
contribute to making SMAPI difficult to use in normal, mainstream life.

1) Building an application to use the SMAPI is fairly complicated  -- the
learning curve is quite steep. RPC programming is not exactly trivial to
understand, and it would be very nice if there was a CMS command-line
utility that allowed a user to exercise some of the less esoteric SMAPI
function without writing a C program to do it. It would definitely make
adoption of SMAPI easier and simpler, and would drive development of the
SMAPI interfaces as *the* way to interact with the system, regardless of the
underlying engine doing the assorted bit-twiddling behind the scenes.

Right now, SMAPI is a pretty neat idea, but it is also pretty much a novelty
item -- after writing a product that attempts to exploit it, I don't think
it will catch on until/unless it is more accessible to the regular CMS user
community. I'd really like to see SMAPI incorporated into the CSL, with the
accompanying language bindings to make it more accessible, and the
abovementioned CMS utility to drive some of the more useful features. It
would also remove the prereq of licensing a C compiler just to use the
SMAPI.

Another way to think about it is to compare it to the impact that RXSOCKET
had on the use of VM TCPIP. You released a perfectly good C socket API
library -- but how many people really exploited it until Arty Ecock loosed
his super-convenient way to exploit it on the world? Maybe someone needs to
pry Arty away from his shiny new z990 toy and get him to produce a
RXSMAPI...8-)

2) Distributing precompiled modules created with LE is still a pain because
of the requirement for the runtime library for C. Yes, I know about the
included LE runtime, but I still find it a support mess to deal with
multiple possible library levels at runtime and I still find occasional
differences in the licensed LE libraries and the runtime LE supplied with
z/VM. That you can't really fix (at least without taking over compiler
development for VM and reversing a long-standing trend), but it makes
exploiting function like SMAPI harder than necessary.

Before you say it, yes I've sent in comment cards on #2. #1 is more a
enhancement request so that others need not go through what we did to get a
handle on how SMAPI works.

Reply via email to