louis@auvax.UUCP (07/14/87)
We are starting to plan for wide-area networking of PCs with our
central UNIX machines. The PCs will be spread out across the Province
of Alberta with dial-in access and/or Datapac (X.25) access to the
central machines. At the central site we now have Berkeley UNIX on
VAXEN and SUNs connected by Ethernet, but we will also be considering
3B2s. The PCs will be used by students to run CAI courseware on MS/DOS
but with administration and tutorial services from the central UNIX
systems. Some CAI will also be delivered directly from the UNIX
systems. We will be starting with 1200 and 2400 baud lines. The
PCs will be standalone workstations with only a modicum of computer
skills assumed for a typical student.
What experience has anyone had using TCP-IP on PCs for wide-area
networking?
Where can we get information on SLIP?
Any other suggestions?
Please e-mail me and if there is enough response I will summarize.
--
Louis Schmittroth My employer has no opinions.
Computer Science
Athabasca University ...{ubc-vision, ihnp4}!alberta!auvax!louisccruss@ucdavis.UUCP (Russ Hobby) (07/15/87)
SLIP connections seem to be a hot topic this summer so I thought I would
post a description of a summer project we have at UC Davis. SLIP is a
cheap and easy method to get a computer connected to a network and is
particularly good for microcomputers since additional hardware is not
required.
SLIP is lacking in a few areas however. Throughput on slow serial lines
(1200/2400 baud) can be quite bad because of the minimum packet size
(all those header fields). Also there is no standard method of
establishing a SLIP connection for temporary network hookups.
Our project addresses the first problem by using an abbreviated IP
packet on the SLIP line and have the SLIP gateway build legal packets
before sending them onto the network. Likewise, incoming packets will be
reduced before sending them down the SLIP line. Many of the fields in
the IP packet header are either static or unnecessary, IF you consider
the host at the end of the SLIP is a leaf on the network and will not be
routing. Static fields are established at login. Our current plans are
for four or eight byte headers, depending if the to/from address has
changed from the last packet.
To solve the second problem, we are creating a standard logon procedure
that will make the connection and set the static fields. We plan to
support connections via campus serial connection (up to 19.2k) as well
as via dialup modems (300/1200/2400 baud, although I can not imagine it
working well at 300!).
We doing the development with an IBM PC clone on one end and a VAX with
4.3bsd on the other. We have started with some of the MIT PC/IP
derivations on the PC side, so that if the project works out, it will
usable by others using the same software. If there is interest, I will
post information on how well it works.
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.edu