[Info-vax] BASIC compiler in the hobbyist distribution
Stephen Hoffman
seaohveh at hoffmanlabs.invalid
Sat May 16 12:06:42 EDT 2015
On 2015-05-16 15:15:39 +0000, John Reagan said:
> I suspect the defaults of "NO" for the "Full Install" was based back in
> the days of TU58s or TK50s. Reading the next saveset takes time....
> Today, it doesn't make much difference with DVDs.
The variant installations date back to the era when everybody had slow
systems, tiny disks and glacial tapes.
> I doubt anybody uses the BASIC Translator anymore.
Y'all really should work on system configuration and usage survey tools
— intelligence-gathering — as bad data or no data causes more problems
than it should.
> One of my plans is to switch the remaining compilers that use VMSINSTAL
> (BASIC and Pascal) over to using PCSI to be consistent with the other
> compilers (I don't want to start another discussion on whether PCSI is
> good/bad/modern/etc.). I'll change the defaults and/or just eliminate
> some of the questions.
Until y'all either overhaul PCSI or replace it with something more
modern, PCSI works well enough, and is usually preferable to using
VMSINSTAL.
FWIW, as for options and alternatives and variant installations and
questions, the more you can remove all of that dreck and provide one
installation and one option and one configuration, the better. The
"pick your directory" dchtick and the "pick which subset(s) get
installed" are among the worst to deal with, as those options increase
the product complexity and the testing effort and the errors for all
sorts of code elsewhere in the product — both for VSI and for those of
us with application dependencies — as that detection and that variation
needs to be customized to figure out where stuff was installed, and
just what is installed. How many times did < and > delimiters get
mishandled somewhere, after all?
If you really think that there's a need to allow the layered products
and the application-private data somewhere other than the system disk,
then architect an OpenVMS-wide mechanism that is available across and
works consistently across all of the products and all of the tools —
maybe add a third system root in SYS$SYSROOT, for instance — and update
all the tools to deal with this stuff in a consistent and sane and
simple fashion. The existing ad-hoc per-product customization approach
is crappy for everybody involved.
Corollary: the more you can roll into the base distro, the better.
Apache, for instance.
ps: addressing the configuration and survey and the accessibility of
the data and the limitations of PCSI, having a way to find out what's
installed would be handy. A PCSI callable API of some ilk. Both
OpenVMS Engineering Nashua and OpenVMS Engineering Bengaluru got into
problems with variations of this mess and things would have gotten more
gnarly as the use of this continuous-deployment practice accelerated.
This both by using the same GSMATCH across all versions of the various
language RTLs, and also by releasing software enhancements via patch
kits, given that there was no (supported) way to see what enhancements
and features were available — what UPDATE kits were installed — at
run-time. The old workarounds removed the "blade guards" against
downgrades, but the problems that those prevented still exist. You're
probably going to be headed toward continuous deployment sooner or
later, so we're all going to need to have a way to figure out what
widgets are present short of running and crashing, or digging around in
API definitions or similar — when the deployed version number doesn't
differentiate the system. Think of what will happen with applications
that target Windows 10 over time, given Microsoft are planning to be
doing rolling or continuous deployments, for instance. The PCSI API
would obviously be useful for configuration collection and application
and system crash analytics, too.
--
Pure Personal Opinion | HoffmanLabs LLC
More information about the Info-vax
mailing list