[Info-vax] Rethinking DECNET ?

[Johnny Billquist]

On 2014-09-06 16:07, johnwallace4 at yahoo.co.uk wrote:
> On Saturday, 6 September 2014 12:05:54 UTC+1, Johnny Billquist  wrote:
> [snip]
>
> "OSI, in contrast, was a committee deisgn"
>
> What do you see as the difference between an RFC and a committee design?

The RFCs as such are just the formalization (if you can call them that) 
of something thought up by a few people, or even just one individual. At 
some point a person or a small group have worked out something to the 
point where they think it is good enough, and stable. At which point you 
write the RFC in order to have this documented, so that other 
implementations can be done, which, if they follow that RFC, should 
interoperate just fine.

> "very few full implementations around (did really any other that DEC do it?)"
>
> Multiple other vendors had connectivity and interoperability, but may
> not have had full integration with their own network stacks.

If they didn't have full integration with their own stack, what did they 
have then?

> "in the 80s, TCP/IP proved that it already existed, it already worked, and
> was already available on multiple platforms."
>
> We may have to disagree, especially if you mean mid 1980s.

Certainly mid 80s.

RFC 801 is dated November 1981 
(http://www.rfc-editor.org/rfc/rfc801.txt), and there you can see the 
implementation status of TCP/IP on various systems at that time, and 
even VAX/VMS is in there.

The you have RFCs like 846 "Who Talks TCP? -- Survey of 22 February 
1983" (http://www.rfc-editor.org/rfc/rfc846.txt), which is just a 
checking of accessible machines at that moment. The answer is several 
hundred, which was quite significant for something in the beginning of 1983.

So yes, by the mid 80s, TCP/IP had already proved itself.

> "[OSI] interoperability was still not proven"
>
> Says who? In the mid to late 1980s and beyond, there were standards,
> there were the equivalent of "plugfests", there were conformance test
> suites. Conformance does not of itself prove interoperability but it's
> a pretty good start.

Yes. But mid to late 80s are much later than TCP/IP, which had already 
proved itself before then. So I definitely stand by that statement. OSI 
interoperability was definitely not proven at the point in time when it 
was relevant.

> "in the end, people took what worked."
>
> Short term people took the cheapest short term fix. There are a lot of
> cheapskates around.

Makes no sense to go for something sluggish, expensive and barely 
working, when something quick and well working already existed.

But I know that many computer companies would have preferred that path, 
as it would have led to more sales of software, hardware, and 
specialized services...

> "Microsoft was not even on the horizon back in those days."
>
> Microsoft were admittedly a little bit late "getting" the Internet.

That was an understatement. :-)

> GOSIP was early 1990s. WinNT was early 1990s.

Anything that goes into the 90s are too late to the table to do anything 
else than just eat what was being served. By then, OSI was already dead. 
It just took a while for some to accept it.

	Johnny