[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [MiNT] mint web/apache idea



Hi,

On Thu, May 11, 2000 at 10:07:45AM +0200, Jo Even Skarstein wrote:
> > I don't see a way to perform parallel network transmissions with the
> 
> It's not possible...
> 
> > current API.  But pseudo-threaded network io is a must for a web browser.
> 
> ...but you can cheat a bit:
> 
> 1. Translate all <IMG>-tags to <EMBED> in CAB.OVL. You'll
>    have to do some magic to the filepaths so it looks like
>    a local file while still keeping the URL. Otherwise
>    CAB will download it itself.
> 2. Write OLGA/ID4-clients to download and display gifs and
>    jpegs. It can be a multithreaded client or you can spawn
>    new apps for each image. You can also load and use
>    CAB.OVL if you like. Oh, it must also handle imagemaps.
>    And it's own cache...
> 
> With this relatively easy hack CAB itself will only download
> the HTML (and the 0.001% PNG-images) while the images are
> downloaded and displayed in the background. It'll be a real
> memory-hog though.

Good idea but far too complicated.  If you write a w3c-libwww based ovl
you can do that better.  The library has a crude HTML parser (sufficient
for filtering out images, sound, video and so on).  You can simply
prefetch these files.  The library also comes along with its own cache
routines (both volatile and persistent) so that you have probably already
started downloading the embedded objects when CAB requests them.

I've started such a project two years or so ago and it was buried with one
of my broken harddisks.  But it already worked (although I probably wanted
too much then) in a basic way.

Today I wouldn't start such a project again because CAB development is
dropped and CAB's html parser is already behind.  I don't see a point
trying to improve CAB.

> > Except for the GEM programming, writing a non-proprietary 
> 
> I could probably help out a bit there, GEM is the only thing
> I master...
> 
> web browser for
> > Atari only needs a team of volunteers.  It's simply too much for a single
> > person but for a team it should be possible.
> 
> Even more important than volunteers is the coordinator.
> Without a bomb-proof design and a skilled programmer to
> coordinate the efforts such a project is doomed.

And the name of this coordinator will most certainly not be Guido. ;-) :-(

I neither have the time nor the skills nor the resources to master such a
project, sorry.

Another prerequisite for such a project (I know that from my Sparemint
experience) is an internet site that hosts the development.

OK, who wants to be the coordinator/maintainer for FAB (Fabulous/Free
Atari Browser)?  Joska volunteered to contribute to the ftk (the FAB
Toolkit, you don't want to write a web browser without some kind of
toolkit), to a certain extent I can give some support for the networking
layer (i. e. interface to w3c-libwww which I know a little).  Josephus
also offered help in a private mail (I hope he won't mind seeing his name
mentioned).  Now go ahead for the rest, Atarians!  Who is going to take
the crown?  And who is going to find a better name than FAB?

Ciao

Guido

P.S.: If the browser is based on the w3c-libwww, then it will actually not
be an Atari browser but rather a MiNT browser.  In fact, every network
application has to decide on some TCP/IP layer and I think that on this
mailing list it is probably agreed that this will be standard BSD sockets,
i. e. MiNTNet.
-- 
http://www.stud.uni-saarland.de/
Send your spam to president@whitehouse.gov and your replies to
mailto:guido at freemint dot de