 Re: strange tcpip issue
On Aug 25, 10:44 pm, Rich Jordan wrote:
> On Aug 25, 1:20 pm, JF Mezei wrote:
>
>
>
> > Rich Jordan wrote:
> > > It was the government weaseling out of the POSIX and OSI mandates that
> > > pulled the rug out from under DEC and the other folks that had
> > > bothered to implement it.
>
> > I am not sure "weaseling" is the correct word. TCPIP grew phenomenally
> > because all of a sudden, Unix started to be taken seriously. (Sun may
> > have had a lot to do with it, stealing DEC customers and moving them to
> > TCPIP based Unix).
>
> > I think that the US government maintained the OSI mandate long enough
> > for DEC (and I think HP and IBM) to implement their stack and then
> > admitted that TCPIP had become the de-facto standard that allowed
> > computers from any manufacturer to talk to each other (the primary
> > purpose of OSI).
>
> > The governmments (this includes europe as well) had a vision of a
> > neutral stack (OSI). Something commercial was developped, but meanwhile
> > TCPIP came along at a much faster pace and responded to the needs of a
> > neutral networking platform.
>
> > You also need to look at Cisco. They came out with a gizmo called a
> > "router" which allowed simple boxes to do that routing job, which made
> > it possible to have simple TCPIP nodes without needing the equivalent of
> > "DECnet routing". And that gave TCPIP a big push.
>
> JF,
> I'm working on memories without hardcopy here, but we had
> meetings with company management and DOE suits, seminars, boxes of
> documentation about the "mandatory" use of OSI, POSIX, (GOSIP???) for
> any and all government contracts after a certain point.
>
> Our company did a fair amount of research, some of the IT group
> was working with DEC (and possibly Sun, we had those too) on OSI early
> testing, gotchas, etc. I know there were many many man hours spent in
> my group (which was not that big).
>
> Then suddenly 'never mind' use TCPIP.
>
> I'm not saying it was a good or bad decision overall, but it was
> a suckerpunch to people who had been told in no uncertain terms 'you
> will do this if you want to keep working with and selling stuff to
> us'.
>
> There's an element of 'weasel' there.
>
> Rich
GOSIP = Government OSI Profile. There were different OSI "profiles"
depending on required application environment. For example, a desktop
environment might follow the TOP (aka Technical and Office) one, and
manufacturing automation networks might follow the MAP one. There
would be conformance tests to ensure that anything claiming
conformance was actually vaguely capable of doing the job, and maybe
there'd even be interoperability tests too, just to make sure that end
users and vendors understood that conformance .ne. interoperability.
As you point out, big vendors of the time such as DEC, IBM, and Sun
(and/or systems houses on those platforms) did put a fair amount of
effort in to supporting this stuff by the mid 1980s, together with
players in specific fields (for example, in the automation sector,
Modicon and Siemens and GE Fanuc and others). There was one big name
in the computer world who weren't (afaik) playing along, but back in
those days their volume OS was still stuck in the 8086 era with a 640K
address space (maybe more so long as you didn't mind jumping through
various hoops) and didn't really do networking except as an
afterthought; it would be 1993 before they had the first release of a
proper 32bit OS with something resembling proper networking, and
another few more years again before NT caught on in volume. NT was
originally going to be POSIX compliant but iirc eventually they got
"grandfathered" out of the POSIX requirement somehow, and the OSI
requirement probably went the same convenient way.
Such is the way that IT history is made (and is repeated, just look at
the current situation with open document formats...).
 |