rdhobby@ucdavis.edu (Russ Hobby) (12/11/87)
Here is what happened at the TCP/IP Interoperability Conference with
regards to SLIP.
The BOF session was too large to get any protocol standards work done,
but the group provided input as to what needs to be covered by the the
standard and some of the potential uses for SLIP connections. There
seemed to be two types of usage for the new standard SLIP protocol. One
is the connection of isolated computers that do not have access to a
LAN. PCs would be a big proportion of the computers in this category,
but certainly any type of computer can use the capability.
The other use of SLIP connections is for temporary network links. The
idea is to advertise a route to another location all the time, but to
dialup and establish the connection to that location only when there is
traffic. The first packet will be delayed while the connection is being
made, but there after traffic would flow smoothly. After a certain
period of no traffic the connection would be dropped, thus saving on
communications costs.
The evening after the BOF session, a smaller group of us got together
and got some productive work done towards the writing of an RFC for a
SLIP type protocol. Here is an overview of what was decided (at least
what I can decipher from my notes).
1) The RFC will cover both connection and line protocols
2) Connection specifications will cover negotiation of items such
as network addresses, authentication, line speed, compression
used, and probably other items.
3) The line protocol will contain one byte in the header to
indicate the protocol in the packet. This allows the use of line
control packets for maintaining the serial link, as well as
allowing other protocols in addition to IP and the line control
to use the connection.
Much more detail was discussed, but I don't want to bore you with it
now. It was thought that an RFC draft could be done and an
implementation to test it in a couple of months.
UC Davis will be implementing the new standard by modifying our current
work. So some of you may want to wait for the standard version rather
than using what we have available now. But then if you have an immediate
need, out old version will be there. Since the standard version will be
more than experimental software, it will more complete and better
documented.
Russell Hobby
Data Communications Manager
U. C. Davis
Computing Services BITNET: RDHOBBY@UCDAVIS
Davis Ca 95616 UUCP: ...!ucbvax!ucdavis!rdhobby
(916) 752-0236 INTERNET: rdhobby@ucdavis.edumminnich@UDEL.EDU (Mike Minnich) (12/12/87)
Isn't there a problem with dynamically changing the kernel's routing table (via a slattach/sldetach mechanism) while user level routing agents such as egpup, routed, or gated are running? This has typically not been a problem for hardwired SLIP connections, since the slip interfaces can be configured before routing agents start up, and are left alone thereafter. One solution would be to home all dial-in connections on another host that would presumably use static routing, but that requires the "host" to function as a "gateway", which is something I would like to minimize. mike