On Sun, 13 Feb 2000, Clarence Verge wrote:
> Glenn McCorkle wrote:
> >
> > On Sun, 13 Feb 2000 17:14:51 -0800, Clarence Verge wrote:
> >
> > > I was not using virtual screens so that extra delay was not in the result.
> >
> > > Just doing a page down takes the same time in Arachne as the initial load
> > > - say 5.5 seconds - while Netscape does it in about a third of a second.
> > > That's 18 times faster !
> >
> > > Don't waste all that time correcting broken virtual screens. To speed up
> > > the local interface, don't do ANY more than necessary on a page down/up.
> >
> > I _think_ I see what's happening here.
> > With virtual screens "off"... only the currently visible portion of the
> > page has been "rendered".
> > Doing a page-down then causes the next visible portion of the screen to
> > get rendered. Page-up "re-renders" the first portion and a second
> > page-down re-renders the second portion.
> >
> > IMHO,
> > With no virtual screen to store the entire page in.
> > I know of no other way that it _can_ be done.
>
> Interesting observation and conclusion Glenn;
>
> So, you are saying that Netscape renders the whole DOCUMENT faster than
> Arachne does one (NON-virtual) screen ? It's possible I suppose.
>
Suppose virtual screen is disabled.
AFAIK, when Arachne downloads a page, it preprocesses it. This means it
calculates the layout of text, images an such stuff. That's when it is
saying "Adjusting images, tables and frames..." and "Loading page from disk"
(if I remember it correctly). This is then stored in some temporary files
(or mem). Later if you scroll the page it looks into this file and draws only
what should be visible in current page position. Of course this is slower
than dumping 1 large block from virtual screen to screen, because it
needs to draw lines, images, text lines, rectangles and such things.
Therefore Clarence, we should clear which of these 2 different processes you
understand under "rendering" and want to compare with Netscape.
Peter Ondraska
aceman
