Re: The Final Word on Browsers and the Future

Subir Grewal (grewals@acf2.NYU.EDU)
Sat, 2 Nov 1996 01:37:30 -0500 (EST)


Date: Sat, 2 Nov 1996 01:37:30 -0500 (EST)
From: Subir Grewal <grewals@acf2.NYU.EDU>
To: dsb@cs.duke.edu
Cc: www-html@w3.org
Subject: RE: The Final Word on Browsers and the Future
In-Reply-To: <199610211731.NAA27635@goldfinch.cs.duke.edu>
Message-Id: <Pine.ULT.3.95.961101230428.15809D-100000@acf2.NYU.EDU>

On Mon, 21 Oct 1996 dsb@cs.duke.edu wrote:

:No, he doesn't; on a system like this, he's almost certainly using Lynx,
:which, even on a 286, starts up *faster* than Netscape on a Pentium.

AFAIK no one has successfully compiled Lynx to run on a 286.  A few people
are working on porting Lynx to 386+ machines and there has been some
success.  The Lynx executable is currently too large for most 286
configurations (I'd be pleased to find I'm wrong about this).  DOS-Lynx is
a completely different application and has seen lot less development than
Lynx has (though it is functional).

I suspect what you were trying to say is that Lynx is very fast on the
multi-user system where you have an account cs.duke.edu (presuming that's
where you're running Lynx) appears to be a SUN-SPARC-10, which is I
believe a quite powerful machine.  Yes, Lynx does have the advantage that
most character-cell applications do, that it can be used with very basic
hardware for direct user interface.  This would include 286, Apple IIs,
VT220 terminals (like mine ~:), etc.  But one must remember that Lynx is
actually running elsewhere, it's only displaying on your machine. What's
running on your machine is only a terminal emulator, and they're generally
quite small.  The Lynx binary is over 1MB for our Ultrix and OSF1 systems
at NYU. 

The question of Lynx's speed has various answers which have been debated
on lynx-dev for a while.  These include, execution on powerful multi-user
systems, better connections to the network (i.e. data-lines vs. modem) on
these machines, Lynx's habit of storing the last 10 rendered docs in
memory to maintain ease of movement between them and the lack of GUI
trappings.  When Lynx is run on PCs with Linux (smaller system resources),
it has been known to exhibit memory problems, though I doubt they'd be on
par with the sort of memory hogging Netscape engages in, especially since
Fote spent so much time this summer fixing all the memory leaks there
were. I'm not too sure about this, but if DOS/Windows ports were
available, Lynx's default behaviour of storing 10 rendered docs in memory
might hamper operation, but changing this behaviour would mean rethinking
the user base for Lynx (the defaults can easily be user-configured in
lynx.cfg the token is DEFAULT_CACHE_SIZE, and for a PC-Lynx one could set
up a dummy proxy program that would cache sources on disk).  Similarly, if
a DOS/Windows port is available and Lynx is used by people over modems,
they'll soon discover that Lynx needs to wait for the entire doc to be
parsed before it can display it (though you can z'ap the transfer).  I'd
then expect to see a clamour for Netscape type behaviour (i.e. rendering
before the entire document has been parsed) and from all accounts (Fote
will correct me if I trip over my laces) this will be exceedingly
difficult to accomplish. 

That said, I use Lynx almost exclusively, I think it's a wonderful browser
and wouldn't give it up for anything else I've seen.  I am so confident of
Lynx's strength that I do not see any need to bend terms and facts to
defend it.

hostmaster@trill-home.com * PGP * Blue-Ribbon * Lynx 2.6 * comp.advocacy@NYU
Those who can, do.  Those who can't, simulate.