RSPF vs. OSPF? No holy war, please!
- To: "k1io, FN42jk 04-Jan-1992 0034" <firstname.lastname@example.org>
- Subject: RSPF vs. OSPF? No holy war, please!
- From: ""Milo S. Medin" (NASA ARC NSI Project Office)" <email@example.com>
- Date: Sat, 04 Jan 92 00:11:48 -0800
Well, my point really wasn't one to try and kill RSPF at all. It's just to
point out that any standard protocol (IP or OSI based) will tend to
see a lot of implementations be fielded, and this will spill into a
lot of different environments.
In the Internet too, we have seen folks who want a lower functionality OSPF
for their routers based on PC type hardware. This has caused a group to
try and define a workable subset of OSPF (I call it OSPF lite), which
doesn't support areas, and a lot of the features needed in smaller nets,
but still be interoperable with mainline OSPF routers. There are even
folks who want to go farther, and build a micro-OSPF!
At the end of the meeting where these folks pressed their cases, I asked the
following question: "Isn't the reason why everyone here is pressing for
protocols with smaller memory utilization because of a mistake Intel made about
a dozen years ago with an ugly memory architecture, and that was compounded
by Microsoft building an operating system that was based on that memory
architecture, even though today RAM is cheap, and CPU is available to burn?"
The reply was yes. We should all think twice and remember why we are building
on the platforms we use. Ultimately, the Amateur community will have to deal
with this, just as it moved from a C/PM based environment to a DOS based
one, and will have to move to a still more advanced platform to support
the needs of real networking. It's hard to get kids excited about 1200 baud
half duplex links when they dial local BBS's with 2400 baud full duplex
modems, and 9600 baud modems becoming much cheaper.
OSPF wasn't build for a PC environment with 640K of RAM support. We did
run on 68010 based routers with 512 K of RAM comfortably back in the V1
days however. V2 implementations, and certainly true for stub only implmentatio
could fit in a limited environment if you stripped out all the mandatory
MIB and management support, and used compilers that generated compact code.