[Info-vax] implementing IPv6 on the internet

[Stephen Hoffman]

On 2016-09-23 07:19:15 +0000, Dirk Munk said:

> David Froble wrote:
>> Dirk Munk wrote:
>> 
>>> How we got here is indeed an interesting question. VSI was working on a 
>>> new IP stack, it should have been ready by now. Was that a completely 
>>> new stack, and what happened to it? Or were they negotiating with 
>>> Process Software all the time?
>> 
>> Interesting observation.  You got any personal experience in such 
>> development? Which should include a much better configuration 
>> interface, and much better other stuff.
> 
> Oh no, I have no experience in these matters. I just notice that VSI 
> told us that they were working on a new IP stack, and they even gave us 
> some estimations when to expect it, and now we get the news that they 
> will license Multinet technology. That seems to be a discrepancy.

Why do you think I was pointing to Process as the only way what VSI had 
announced was going to happen?

Why did I make that guess?   IP stacks are large projects, particularly 
if you're writing a wholly new one.   Not as large as writing a 
competitively-featured operating system, but still a large development 
project.   After the initial implementation, there are years of testing 
the interoperation of the IP transport, routing, and various services, 
and the testing of the application code.   You can't implement on the 
specs or the RFCs or such as there are incompatibilities and errors;  
reality divergences from the documentation.   You have to do 
compatibility testing, and more than a little of that.   More than a 
few folks need be working on all this code given the scale of the 
effort, too.    Consider what else VSI has on the schedule around 
Kittson and x86-64 and just sorting out all the rest of the giblets, 
and you'd probably need to double the present VSI staff to add a 
wholly-new IP stack to the development schedule, particularly if you 
wanted the original release schedule that VSI discussed.   That meant 
the IP stack already existed and already worked, and already worked on 
OpenVMS.   Which — since this wasn't marketed as an overhaul or massive 
update of the HPE TCP/IP Services stack — meant that Process had to be 
involved.   Q.E.D.

PR announcements are inherently vaporware.   Pragmatically, call me 
back when we can see the design and the API and the features and maybe 
even the beta, when we really know what we need to change, when we 
known what got better and what (hopefully little) got worse, and what 
the deprecation and removal schedule might be.   What licenses and 
buy-outs were present in the old IP stack didn't matter then, so why 
this now?


-- 
Pure Personal Opinion | HoffmanLabs LLC