No, I don't want it show the rest of the log at all unless explicitly requested;
it will freeze the browser. Even if the tail shows up first, it will soon become
impossible to scroll around just in the tail if megabytes more text are added at
the top. In fact, the 100% CPU usage will likely make even the text currently on
screen not display, in case you switch to a different desktop and back, because
the browser will not get a repaint event. Open
http://deadlock.nbextras.org/hudson/job/javadoc-nbms/292/console (when you have
no unsaved data in your browser!) and see what I mean.
Possible combined UI that would replace the current, somewhat clumsy, pair of
"Console" and "[raw]" links:
1. If the log is small, "Console" displays the regular live scrolling log, with
a "Plain Text" link mainly for those who want to download it.
2. If the log is big, "Console" displays just the tail of the log (in an HTML
page without JavaScript), with both a "Full Text" link for the complete live
scrolling log (if you have a very fast CPU and are feeling lucky), and a "Plain
Text" link for the complete text for download.
Maybe what we need is for the GUI console page to do something really smart,
like just showing the tail portion first then incrementally fetch the rest of
the log backward till it hit the top?