Subject: IPTOS_LOWDELAY in telnet
To: None <>
From: Allen Briggs <>
List: tech-net
Date: 10/17/1999 18:04:21
Should NetBSD's telnet set IPTOS_LOWDELAY?  It seems that with a
telnet connection, you do want a low delay, but you do want some
delay to properly handle multi-byte sequences (like terminal escape
codes).  It appears that both telnet and rlogin set this the same way.

I was thinking about this some in response to someone's comment about
NetBSD telnet not passing escape sequences through correctly all of the
time (i.e., they get the escape characters too far apart for vi, etc.,
to see them as escape sequences).

As far as I can tell on a quick scan, IPTOS_LOWDELAY only affects
ppp*, sl*, strip*, anyone using if_spppsubr.c, and some kind of
serial(?) h/w driver (hd64570.c--what is this?  How could I really
find out except to ask the author?).

if_spppsubr.c actually checks for interactive sessions separately from
those with IPTOS_LOWDELAY set so it seems that in this case, it's moot.
The other cases could probably check this in the same way.

How does this interact with Nagle, or does it?

In another vein, is there any reason (short of a SMOP) why we don't have
gettosbyname()?  Is it a dead-end from a standards point of view?