On Jan 2, 2013, at 7:12 AM, <sampsa at mac.com> wrote: On 2 Jan 2013, at 14:03, Mark Benson <md.benson at gmail.com> wrote: On 2 Jan 2013, at 11:20, sampsa at mac.com wrote: So LAT is as fast over a WAN link as CTERM is over a LAN..What's wrong with CTERM? My understanding is LAT is a self contained ethernet-based protocol purely designed and optimised for terminal connectivity. CTERM is a command terminal application that runs on top of DECnet and works via the DECnet stack. The added layers probably, at least in part, account for the speed variation between the two. I could see that on like Z80s running over a 9.6kbps link, but VAX level gear on a 10 Mbps ethernet LAN? It can't be the protocol overhead alone, CTERM must be seriously brain-damaged by design. Cterm is a pretty awful protocol, "heavy" certainly describes it. Roughly speaking, it defines a remote procedure call style "generic terminal I/O" service. The primitives in the protocol resemble the combination of VMS and TOPS-20 terminal driver features. So if you think of all the things you can ask a terminal to do with VMS QIO$ syscalls, then think about hairy stuff like EDT screen based editing, you roughly have what cterm tries to do. But still, the numbers shown are a bit low. What is the test scenario for cterm? I suspect a part of the problem may be kernel vs. user mode processing. paul