DELETE/BACKSPACE key paradox on RS/6000

[Tolliver J S]

Hi!
I have an RS/6000. Sometimes I log in at the console, where the key marked
Backspace works as intended--it rubs out the previous character. Sometimes,
though, I log in from a VT100 terminal emulator, where the "delete" key
sends an ASCII 177 to the RS/6000, which echoes as ^?. Then one must press
control-H once to rubout the ^? and once more to rubout the original
character one wanted to "delete". This behavior can be fixed with the
stty erase command, but that's a pain. I could put stty erase in my .profile
along with enough smarts to know whether it's on the console or a dumb
terminal, but that seems rather inelegant.

My TERM variable is correctly set to be "vt100". Also, note that this same
behavior results when one telnets to the RS/6000 with a REAL VT100 via some
other host.

I claim, but am willing to be convinced otherwise, that the terminfo entry for
a vt100 is wrong. That it should deal with the ASCII 177 sent by the delete
key on a VT100 (or an emulator) "correctly" and make it behave as the "erase"
character without messing with stty erase in .profile or .login. Is this
reasoning incorrect? Why or why not? If correct, how do I hack the terminfo
entry? I've tried hacking around a bit, but terminfo is hieroglyphics to me.
I found cub1=^H and kbs=^H in the vt100 stanza in the dec.ti terminfo source
code stanza. Interestingly, they were each TWO characters (i.e., a carat (^)
and in H), not real control-H characters. I replaced them with ^?'s, applied
tic, logged out and back in. No luck. Also tried replacing them with actual
ASCII 177 codes entered with the ^V quote mechanism in vi (which echoed as
^?). Still no luck.

Is there a way to do this? How? Is it a good idea? Why or why not?

Many thanks,
Johnny Tolliver
Oak Ridge National Laboratory
jxt at ornl.gov