Mailing List Archive

Re: HotJava servers (fwd)
For those not on the hotjava-interest mailing list.... this is a reply to
my comments on Arthur van Hoff's statement that they were writing an HTTP
server in Java. The last couple sentences are the most interesting.

Brian

---------- Forwarded message ----------
Date: Mon, 3 Apr 1995 13:26:28 +0800
From: James Gosling <jag@scndprsn.Eng.Sun.COM>
To: brian@wired.com
Subject: Re: HotJava servers

> On Sun, 2 Apr 1995, Arthur van Hoff wrote:
> > HotJava does not require a special server. It works with any standard http
> > server. We are working on a Java based http server, but that server is not
> > required to run applets.
>
> This is very interesting - could you elaborate on this? Is the goal just
> to write a basic http server in java because it can be done, or is it to
> implement an extended http to enhance client-server functionality? Or is
> it to give browsers the ability to be servers as well - this is perfectly
> in line with the peer-to-peer model of the WWW (rather than client-server)
> that must come to be if the web will be used for true collaborate
> applications.

We can do a number of things. Some are just implementation nits like
using multithreading instead of Unix fork(). There are a number of
things that fall out from doing this that make for big performance
improvements. But you could do this in C. We also let you use hunks
of Java code instead of cgi-bin scripts, which turns a pile of
fork/exec/script parsing into a method call. A big performance boost.
We can also do application-specific plugins that give the server a
telescript-like behaviour. One little plugin I did implements a simple
"chat room" protocol.
Re: HotJava servers (fwd) [ In reply to ]
In reply to Andrew Wilson who said
> Ultimately the way the web will be used will be defined as much by client
> design as by server design. I was thinking to myself a while back that it'd be nice
> if there was a sister-project to Apache which focussed on building and maintaining
> sources for a SOTA web client. For prettiness, Netscape is still quite good but a
> well ported HotJava-like tool would be more useful to the leading-edge users who've
> inspired the current projects.
>
> If you imagine a super-Apache 3.0 with browsing capabilities then things start to
> get a little heady.
>

How about contacting the arena people?

--
Paul Richards, FreeBSD core team member.
Internet: paul@FreeBSD.org, URL: http://isl.cf.ac.uk/~paul/
Phone: +44 1222 874000 x6646 (work), +44 1222 457651 (home)
Dept. Mechanical Engineering, University of Wales, College Cardiff.
Re: HotJava servers (fwd) [ In reply to ]
> > This is very interesting - could you elaborate on this? Is the goal just
> > to write a basic http server in java because it can be done, or is it to
> > implement an extended http to enhance client-server functionality? Or is
> > it to give browsers the ability to be servers as well - this is perfectly
> > in line with the peer-to-peer model of the WWW (rather than client-server)
> > that must come to be if the web will be used for true collaborate
> > applications.
>
> We can do a number of things. Some are just implementation nits like
> using multithreading instead of Unix fork(). There are a number of
> things that fall out from doing this that make for big performance
> improvements. But you could do this in C. We also let you use hunks
> of Java code instead of cgi-bin scripts, which turns a pile of
> fork/exec/script parsing into a method call. A big performance boost.
> We can also do application-specific plugins that give the server a
> telescript-like behaviour. One little plugin I did implements a simple
> "chat room" protocol.

This is interesting (but perhaps not in the scope of the current round of apache
modifications). Xerox PARC are working on a similar (loosely termed) super-client,
called Jupiter, and based around the MOO OODB-environment, another one of my litle
hobbies. MOOs have already been turned into fully HTTP compliant WWW servers.

Ultimately the way the web will be used will be defined as much by client
design as by server design. I was thinking to myself a while back that it'd be nice
if there was a sister-project to Apache which focussed on building and maintaining
sources for a SOTA web client. For prettiness, Netscape is still quite good but a
well ported HotJava-like tool would be more useful to the leading-edge users who've
inspired the current projects.

If you imagine a super-Apache 3.0 with browsing capabilities then things start to
get a little heady.

It'll probably happen, and if it does then I want some.

Ay.

Andrew Wilson URL: http://www.cm.cf.ac.uk/User/Andrew.Wilson/
Elsevier Science, Oxford Office: +44 0865 843155 Mobile: +44 0589 616144
Re: HotJava servers (fwd) [ In reply to ]
> In reply to Andrew Wilson who said
> > Ultimately the way the web will be used will be defined as much by client
> > design as by server design. I was thinking to myself a while back that it'd be nice
> > if there was a sister-project to Apache which focussed on building and maintaining
> > sources for a SOTA web client.
>
> How about contacting the arena people?

I think we should do that, when Apache is a bit more publicly presentable - it's
working fine but we could manage little more documentation. It think it would be
a good idea to make sure that browsers out there can use the features that apache
provides (content negotiation, last-modified etc.) Let's put that on our list of
things to do for a public release.

> Paul Richards, FreeBSD core team member.

Ay.