> On Jul 15, 2016, at 10:08 AM, Mouse <mo...@rodents-montreal.org> wrote:
>
>> DECnet might be totally integrated and awesome, but it's also
>> proprietary, seldom used,
>
> I think it is only semi-proprietary. I've seen open documentation that
> at the time (I don't think I have it handy now) I thought was
> sufficient to write an independent implementation, both for Ethernet
> and for serial lines.
DECnet is open in the sense that anyone can see or reprint the specs, and
implement the protocols. Arguably it is pretty similar to the BSD license (the
"with attribution" variant). And the specs were written with sufficient care
that following them is, in general, sufficient to create an interoperable
implementation. For example, I implemented DDCMP for RSTS from the DDCMP spec,
and "it just worked". This, by the way, is quite rare in protocol specs; it
certainly is not true for many RFCs, and for one I know of it wasn't even
considered a worthwhile goal by the document editor!
The only ways in which DECnet is proprietary is that the development work was
done by Digital and not others. And the name (DECnet) was a trademark. (Then
again, so is "Linux".)
Actually, the "done by Digital" is true only through Phase III. In Phase IV,
you get Ethernet (developed by Digital, Intel, and Xerox), HDLC (developed by
various telcos based on earlier work by IBM), and perhaps other bits. And of
course, in Phase V, a whole lot of the machinery is from OSI, though that was
very much a two-way street (IS-IS came from Digital's work on Phase V routing,
as did OSPF). Finally, even when one organization did the detail work in a
particular area, various algorithms and inspiration came from other sources.
Dijkstra's algorithm is a good example, of course, but there are plenty of
others. (The softlink loop detection algorithm in DECdns is another example of
a decades old algorithm put to good work in DECnet.)
paul
