Re: xmosaic experience

marca@ncsa.uiuc.edu (Marc Andreessen)
Date: Thu, 25 Feb 93 17:36:10 -0800
From: marca@ncsa.uiuc.edu (Marc Andreessen)
Message-id: <9302260136.AA22163@wintermute.ncsa.uiuc.edu>
To: Bill Janssen <janssen@parc.xerox.com>
Cc: WWW-TALK@nxoc01.cern.ch, ebina@ncsa.uiuc.edu,
        marca@ncsa.uiuc.edu (Marc Andreessen)
Subject: Re: xmosaic experience
In-reply-to: <0fXIpFoB0KGW4v0hoS@holmes.parc.xerox.com>
References: <0fXIpFoB0KGW4v0hoS@holmes.parc.xerox.com>
	<9302252239.AA21530@wintermute.ncsa.uiuc.edu>
	<EfXGE90B0KGWQv0bZM@holmes.parc.xerox.com>
X-Md4-Signature: 1a892af7eddde237559724ae86a443ff
Bill Janssen writes:
> Relax, Marc.  First off, I really like xmosaic, and am hoping version
> 1.0 will be really great!  

It's good to hear that.  Public complaints about something that we
actually put a lot of time into don't encourage us any, though.  I'm
sure you can understand that.

> But secondly, I'm not your customer.  Who cares what I think?

Well, you literally are our customer.  But that's probably beside the
point... we do care what you think simply because having the wonderful
distributed beta team that we essentially have due to this group gives
us the opportunity to make our product much better than it could be
otherwise.  We very much value constructive criticism from
knowledgeable people.

> Well, since you ask...  If you're displaying text, it really doesn't
> matter what the maximum size of a window is, so long as you can get
> at least one line on it, right?  As the user scrolls through the
> text, you just show different lines, right?  But apparently what
> you're doing is turning each document into a giant image of text,
> and then panning through that image, which of course runs you into
> size problems.  This is an OK idea if you're building something like
> HyperCard.  But WWW isn't always like HyperCard, it tends to mix the
> ``holy scroller'' model with the ``card shark'' model of hypertext.
> So this approach isn't really suited either to WWW or to the X
> Window System, so far as I can see.  (The problem with it on X is
> that is uses big chunks of window system resources, typically in one
> big burst that locks out everything else on the screen!)  The
> ``correct'' approach would be more of what Viola does, which is to
> use a text widget.

We are using a widget.  A text widget, in fact -- just one we wrote
ourselves, since the Motif text widget doesn't support the
functionality we need.  We are not using a huge-ass 32K or 64K pixel
high pixmap or image, and we are not consuming memory to store such a
pixmap or anything else 32K or 64K high.  We are using a virtual
window that is that high because that way X handles the scrolling such
that we get the proper expose events at the proper times FOR THE AREA
OF THE WINDOW THAT IS ACTUALLY ONSCREEN.  To do our own scrolling
would require a lot of additional work.  Either someone funds us to do
that work or we are going to do more interesting things with our time.

Spend some time with our code before drawing conclusions, in public no
less, like ``you're turning each document into a giant page of text,
and then panning through that image''.

> By the way, Viola, the only other browser I use, *does* in fact use
> a text widget, and handles my document just fine.

OK, if that's the case then my apologies to Pei.  I have a dim memory
of trying this last year and simply seeing the document overrun the
window size, as happens with Midas -- I'm probably mistaken.

Marc

--
Marc Andreessen
Software Development Group
National Center for Supercomputing Applications
marca@ncsa.uiuc.edu