Re: lynx-dev Re: ^Ve considered harmful

[Bela Lubkin]

Bela>>>> Lynx also works with various OS's native curses, some of which do 
*not*  
Bela>>>> handle this correctly.  So if you use ESC for a quote char, be 
prepared  
Bela>>>> for endless niggling bugs on older systems, slowly dwindling as the  
Bela>>>> years go by...  

Bela>> I know, from my years in SCO Support, that customers always had trouble 
Bela>> writing curses programs that could interpret ESC as a separate command. 
Bela>> It would always delay a whole second before accepting a standalone ESC. 

Tom> There's (in several implementations of curses) a variable ESCDELAY,
Tom> which is used to control that (i.e., the nominal timeout in milliseconds).
Tom> ncurses implements that.

The old libcurses(es) I'm talking about didn't.  The delay was in fact
implemented with either sleep(1) or select() with a timeval equal to (1
sec, 0 microsec), and nothing about it was user tunable.  Bad
implementation?  Certainly, but it's out there.

Tom> Of course vi (and clones) use ESC.  Recently I was puzzled about a problem
Tom> with the ESC key before I realized that I was getting extra delay time
Tom> because I was running in 'screen' (I'll revisit that though since screen's
Tom> delay was a factor of ten higher than I expect - but at least vile was
Tom> working properly ;-).

... but this is a fine example of the sorts of things we'll be dealing
with forever, if we make ESC a command character in Lynx.  In many cases
they'll turn out *not* to be bugs in Lynx itself, but will be perceived
that way.  Just like ^V has turned out to be somewhat problematic
because some systems consume it as lnext, others do not, and control is
largely out of Lynx's hands.

>Bela<