david@pyr.gatech.EDU (David Brown) (03/14/89)
Hi. I'm interested in using an Encore Annex Terminal Server instead of regular Sun terminal multiplexors on a Sun server. It seams that they would offload a lot of the I/O interrupts (line editting, etc) so the Sun could perform a lot better. Anybody out there had any experience with Annex terminal servers with non-Encore hardware, especially Suns? Any responses greatly appreciated. --David Brown ---------------------------------------------------------------------------- David Brown Armstrong State College, Savannah, Georgia uucp: ...!{akgua,allegra,amd,hplabs,uunet,seismo,ut-ngp}!gatech!gitpyr!david ARPA: david@pyr.gatech.edu
lamy@ai.utoronto.ca (Jean-Francois Lamy) (03/15/89)
We have one and like it. We have 24 lines coming in from a departmental switch, a pair of TrailBlazers, a Hayes and outgoing lines back to the departmental switch. We do UUCP over the TB modems. We are running version 4.0 of the software. You can run it with security disabled, which is only a good idea if you can trust anyone that will ever reach the annex. We can't, so we run with security enabled. This means our users have to log on the Annex, information gets encrypted and sent to a responding authentication server for validation. They then get to pick whatever machine your heart desires. Even though authentication has taken place, you don't want the annex in the hosts.equiv file because of a bug in the way they implemented rlogin -l (Xylogics says fixed in 4.1), and even if that gets fixed, if for whatever reason a user gets a port that is still active you don't want him to walk into the previous user's account. So our users have to give their password a second time. I'm not aware of complaints about that aspect of the procedure. We had ALM-2s and tossed them (they were starting to run almost right a patch release 5.3, but our patience had run out by then -- we were the first customers in Canada to go to 4.0, before Sun Canada did, because ALM-2s were supposed to work :-). The terminal server is better suited to the way our lab is built, with small research groups getting more and more specialized or dedicated hardware, and where sticking serial line muxes in all those machines would make little sense. Glitches with them include - can't fit the monster routing tables for NSFNET under 4.0. Fixed in 4.1, does not matter if you reach it via a default route. - rlogin -l with security enabled is a security hole, but not if security is disabled (!) - There does not seem to be a way to get them to send a break to the remote computer and, going the other way, to get "tip" to push a break to the outgoing modem. Good things: - They work! - Our users did not complain about anything. Emacs runs fine, ^S and ^Q go through properly, and all that. - They support name servers. - More parameters than you'd ever wish for, with sensible defaults. Jean-Francois Lamy lamy@ai.utoronto.ca, uunet!ai.utoronto.ca!lamy AI Group, Department of Computer Science, University of Toronto, Canada M5S 1A4