[Info-vax] CRTL and RMS vs SSIO

[Stephen Hoffman]

On 2021-10-14 00:36:58 +0000, Lawrence D’Oliveiro said:

> Or, to put it another way, how many of them would make the job of 
> porting to new architectures better, instead of worse?
> 
> Answer: none.
> 
> Except Linux.

At this point in the VSI port to x86-64, your suggestion to re-host to 
Linux is worse.

Short version:

the OpenVMS x86-64 kernel is already available.

Longer version; why re-hosting to the Linux kernel is worse:

Going to the effort of swapping a ~forty year old monolithic kernel for 
a ~thirty year old monolithic kernel while preparing for next-decade 
system hardware architecture designs seems... unwise.

The this-decade hardware architecture details are already here, or are 
arriving soonest.

Unwise particularly given a working OpenVMS x86-64 kernel is available. 
 Getting LLVM working native is one of the larger parts of the 
remaining OpenVMS release schedule, too. Which isn't a kernel issue, 
and won't benefit from re-hosting the kernel.

Unwise given the IP lawyers necessarily get involved with discussions 
re-using a GPL kernel and/or re-using drivers for a closed-source 
product and particularly where the vendor reportedly has no rights to 
release much of the closed source.

As for the portability of BSD kernel designs, the major platforms 
already support all of the processor architectures that people running 
servers are interested in now, as well as Arm and RISC-V:
https://www.freebsd.org/platforms/
https://www.netbsd.org/ports/
https://www.openbsd.org/plat.html

The BSD licenses also tend to lack the GPL-related compliance issues 
that can arise for closed-source vendors.

For folks that do want OpenVMS-flavored APIs on Linux, that's already 
available from Sector7. That product specifically targets migrating to 
platform-native APIs, as well.

For folks that want access to Linux, BSD, Windows Server, or otherwise 
on x86-64 and with OpenVMS on x86-64, hypervisor guests are already an 
option.

(It'd be interesting to see if a hypervisor could add Galaxy-like 
communications features, as the hypervisor and what SRM provided Galaxy 
aren't all that different, but that's fodder for another thread.)

Would re-hosting the existing kernel to a more modern kernel be 
interesting? Sure. Maybe. Eventually. But not prior to V10.0, given the 
available kernel. And not onto the current Linux kernel.

Entirely FWIW and for those interested in "newer" or "different" 
operating system kernel designs from the ~recent past—and to be 
absolutely clear, OpenVMS will ~never see these used—there's the 
~thirty year old Grasshopper design, and the always-weird Singularity 
design, and the VMS-friendly DEC MICA design:
https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.855.3247&rep=rep1&type=pdf 

https://www.microsoft.com/en-us/research/project/singularity/
https://en.wikipedia.org/wiki/DEC_PRISM#MICA


-- 
Pure Personal Opinion | HoffmanLabs LLC