jqj@HOGG.CC.UOREGON.EDU (04/27/89)
The discussion of acceptable delays is rapidly moving away from topics of interest to the majority of TCP-IP mailing list readers. Perhaps the discussion could be moved to another forum? Three technical points: 1/ It is generally agreed that miminum variance in response time is as important to user comfort as mean response time. Studies have been done varying the response time of commands in "line mode" i.e. from the time you press RETURN to the time the command completes. I imagine that there is literature investigating variance in full duplex echo delay. User expectations also make a big difference as to what is considered acceptable. 2/ DEC in their design of the LAT protocol invested substantial effort in literature review and some human factors testing. I believed they concluded that it was acceptable to collect typeahead for 80ms from a terminal before packetizing and sending it to a host, resulting in echo delays to a lightly loaded VMS system with a typical range of (I'm guessing) 40 to 200ms. The 80ms is a tunable parameter. Maybe someone more familiar than I am with LAT could comment further. 3/ A request: could those of you who refer to human factors literature please provide citations? A good though now dated survey is Ben Schneiderman, "Response Time and Display Rate in Human Performance with Computers," ACM COMPUTING SURVEYS, v 16 n 3 (September 1984), pp. 265-286. JQ Johnson voice: 503-686-4394 Director of Network Services Internet: jqj@oregon.uoregon.edu Computing Center Bitnet: jqj@oregon University of Oregon UUCP: ...!uoregon!jqj Eugene, OR 97403 (Internet is preferred)