[Info-vax] A possible platform for VMS?

[dodecahedron99 at gmail.com]

On Sunday, March 1, 2015 at 5:46:42 AM UTC+11, Stephen Hoffman wrote:
> On 2015-02-28 16:04:32 +0000, Dirk Munk said:
> 
> > Stephen Hoffman wrote:
> >> On 2015-02-27 10:27:44 +0000, Dirk Munk said:
> >> 

<snip>


> But getting VMS into the ATM market?   Or into Point-of-Sale systems?   
> Or onto NUC-sized servers?   Or onto high-end systems with hundreds of 
> cores, for that matter?  Or displacing existing applications and 
> services that are running Windows Embedded or iOS or one of the Linux 
> or BSD variants, or some HPTC-focused distro?   That's no small 
> investment, and that puts VSI into a price and feature competition with 
> some large and established vendors, and with some entrants building on 
> Android and iOS platforms; with lower-cost platforms and what is likely 
> lower-cost software.
> 
> VSI is going to need to engage resellers that foresee being able to 
> make a profit here, as VSI itself is probably not going to be creating 
> or porting the software necessarily involved in most any of the areas 
> I've mentioned here.  Tossing a NUC running VMS over the wall might be 
> fun, but there's a whole lot more information and a whole lot more work 
> -- prices, features, support, compatibility, etc -- needed before that 
> might ever become a viable product.
> 
> Getting VMS over onto x86-64 will solve problems that some existing 
> OpenVMS users have.   The rest comes later.  This all assuming that VSI 
> can keep going, and can start to grow the VMS installed base, and start 
> to show that there are and can be revenues for third-party software and 
> hardware providers, too.
> 
> -- 
> Pure Personal Opinion | HoffmanLabs LLC

Agreed

Most of the discussions here are straddling the notion that vms needs to be profitable soon and that means targeting existing customers and the need for vms to modernize to open up new avenues

On the second front, is why I like the notion of pushing into the HPC arena. Unlike other markets that are already saturated, the HPC market is still being defined, especially on the large data store side of things with even linux being reworked to accommodate the exabyte camp. How feasible is this market to target? I have no idea but I do know there's a number of players clambering to get there because it's still being defined

If I could wave a magic wand and morph vms into anything I wanted, I would have it expand way further the concept of galaxy into a Multiverse, under which you have universes and galaxies and superclusters etc in a dynamic formation but centrally controlled if that's what a customer wants.

I would expand the concept of the uic or object protection so that it can start at the top of the chain listed above all the way down. In file headers I would embed encryption that could lock the contents of files or parts of files by date for release, or limit access by way of contact with a central security server if that's how a customer wanted to configure their multiverse.

I'd start by looking at the multiverse as just that, a top down approach where the whole of say a government would be a multiverse (think say the UK government), under which entire universes of vms systems sit as an arm of government such as defense, then further drill down to a galaxy which I would redesign as a super cluster and so on and so on all the way down to the individual node

Other systems now are continually pushing upwards trying to wrap up management and security under ever higher levels of centralized management, so why not just extrapolate all the way up and start with a design right up as high as you can go and then push downwards - someone had the concept of a cluster many years ago, why not go further all the way up to the multiverse?

I know it's fanciful and lofty minded but without a long term vision of something that is different to what other systems are even dreaming about (I have not seen anyone proposing such madness, the furthermost I have seen is fully distributed systems) vms will be forced to compete against other systems who have already cornered their particular markets

With a top down design as lofty as this, you can bring higher security lock down into play because you have a more unified model

Here's some of the idea I have been thinking about (don't prod and poke at it too much, I have not spent a lot of time totally thinking it through) but uic's would become entities like: [multiverseid, universe, galaxy, solar-system, super cluster, cluster, node, group, member] etc. Like IP6, you can short form parts of it and wildcard other parts. uic ownership is inherited from the multiverse down, it cannot be assigned upwards beyond what the security server dictates. In this way, a document produced by say a certain arm of government can set a policy to never allow the uic to be granted to say a public node of say the social services department etc. The key is of course a security server, which could be a universe in and unto itself
I believe the "THE" OS had a notion of giving permissions on a document that could never in turn be granted to another even if copied several times via other parties.

I'd also like to see things like the sys areas completely locked down with staging areas where patches are installed to and moved to various phases before being accepted into their final SYS area. You could have these areas checksummed and verified against say a code base at vsi or something to ensure your system is always running in a known code base configuration. VMS system areas get into a royal mess over time. Why we are at it, how about vendor code also being segregated into areas that totally isolate them from the OS. Vendors could register with say vsi and be given a code id or a unique naming convention that would then be applied across all vms systems out there in the world. Vendors could then install into these areas and could in turn be verified as a base install using a checksum arrangement to ensure code is valid. You could then checksum across base os installs and vendor installs to create higher and higher levels of checksum authentication. I was looking into perfect checksum algorithms but the maths was well beyond me :-(

Once you have checksummed base systems and vendor code, it becomes easier to validate and catch a rougue exe from running. If the exe isn't registered in a known directory for execution, then it is thrown out. This to me is how production systems should be running, code verified and registered before running. In production you should work with known code bases only

Yes, vms needs to be viable again or it's gone :-(  but we need think tanks for future ideas of where we can take vms so as to have a vision to work towards or the linux's of this world will continue to eat our lunch

Sorry for being so dreamy about vms but I'm going to steal out of context, a biblical quote that I think is apt here, "Without a vision, the people perish"

Just what is vms's long term vision? I don't think it's just up to vsi to tell us (even if they have the ultimate say), it's up to all of us to put forward where we believe vms should/could go. We start with dialogue no matter how crazy the idea, then thrash out what can and cannot be done, but without an overall plan of where we want to end up, were going to get sidetracked into little alcoves along the way

My vision for vms's ultimate destination is the multiverse concept, don't laugh please but instead put forward your ideas