[Info-vax] New VSI Roadmap (yipee!)

[lists at openmailbox.org]

On Thu, 26 Feb 2015 11:09:54 -0800 (PST)
dodecahedron99--- via Info-vax <info-vax at rbnsn.com> wrote:

> Rapid development comes down to libraries and frameworks, at least for
> new stuff

I didn't mean to say the focus should be on libraries or frameworks, only
that if the language is already correctly implemented then it seems
wasteful and in practice is usually destructive to continue monkeying with
it. The language really just needs to be standards compliant (in other
words, leave the language design to committees of language lawyers who
are usually very good at that) and just do bug fixes to the language. All
the attention should be directed at keeping the libraries up to date with
OS features and API changes (hopefully not many of those). It's a lot safer
and saner to extend the language via runtime (i.e. callable services) than
it is to add constructs and extensions to the base language. If you're not
careful with extensions you can paint yourself into a corner where people
are unwilling or unable to port code in that language back and forth from
your platform and others. If you're the only game in town and some OS have
their niche then this really doesn't matter. If you want to compete with
commodity crapware OS then you really need to stay close to the standard.
The biggest proof of this is gcc which has become its own language and is
abused to the point stuff written in gcc is no longer portable. They've
blotted out most of the native species so this is less of a practical
problem today. But tomorrow when they pull the rug out and go GPL on their
runtime they will have conquered the world because people were too busy
licking up the freebies to pay attention to what they were really paying.

> My vote goes towards implementing something like MS's .net framework
> where they created their CIL layer for which all languages compile down
> to as an intermediary language. This can then be shipped to run anywhere,
> theoretically, independent of underlining Os version and hardware
> architecture 

I don't know anything about that so I can't comment on it. But in general I
think the approach of closely coupled hardware and software (MVS, VMS) is
better for quality and performance than abstraction as a fundamental
approach (UNIX, Linux, etc.) and as a developer I personally prefer to work
on the former and not on the latter. It seems to me there are so many
issues that arise when going the abstraction and portability uberalles
route like poor performance, complexity creep, increased code path, more
difficult problem diagnosis...the list goes on and on. It's really not
worth it.

It's surprising to read your comments about Microsoft given how they really
only run (arguably) on one platform. I don't know why they went to the
expense given how little they seem to exploit all that for anything useful
other than a common runtime for various languages- and they really don't
have many of those either. I guess you can get away with that when you get
everybody to sell your "OS" with every new PC and server.

> Having some type of CIL code in VMS would free us (or go a long way
> towards) allowing us to change architectures (I'm thinking, Application
> code here) more easily going forward without the need to always recode,
> recompile, retest over and over again when VMS moves onto a new
> architecture. This whole recoding cycle is why, where I work, VMS is left
> to die in the back corner running an application that they refuse to
> modernise because of the huge recoding and retesting effort involved just
> because HP insisted on loving Itanium for too long

I have no experience with what you wrote but since you are talking
about application code why can't you just recompile the source on the new
target platform? This is exactly why standardized languages and
implementations that adhere to those are created.

> Running an existing application inside an emulated environment (charon
> etc) to escape hardware obsolescence is ok for a while but as the OS
> moves onwards and changes architectures then having to maintain several
> versions of an emulator will eventually wind us back in the same spot we
> are in today with the Vax and Alpha and Itanium architectures. Having
> code compile down to something like CIL will (should) make porting an
> application to a totally new architecture much easier as your not having
> to test for functional hardware changes as much

But the performance hit and complexity make it not worth it most of the
time. And this is part of the reason you could run Windows NT in 16M of RAM
on a 25 MHz CPU and why you need 8GB on a 3.2GHz dualie for a Windows 7 box
to give you the same level of responsiveness. If you're in the business of
selling hardware that could be ok. I think the whole thing about VMS
displacing Linux is never going to happen. I would hate to see VMS try to
support POSIX and gnu-crapware and I believe if they do make that their
mission they'll be out of business in 5 years. I would much rather see VMS
capitalize on core strengths of sanity, somewhat miserly resource
consumption, availability, and close coupling of hardware and OS. And good
doc! There are few Unices left with good doc and Linux doesn't show up on
the radar at all.