[lwip-users] Re: [lwip] The future of lwIP

[Paul Sheer]

------=_Part_1191_3913634.1029844245506
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit


Hi Duncan

PaulOS is not going to be publically released just
yet. However, anyone is welcome to look at the code,
provided you are going to help me make it better or
port it to some new platform. I am mostly interested
in people who are going to audit the code or verify
its robustness. If they put in enough work, they will
be allowed to use it in their own commercial projects.

I am not interested in novice colledge students
looking at the code for projects whose only audience
will be their supervisors. This is serious stuff, and
most people don't know how difficult it is to write
something that really is completely solid.

What I am talking about are the many weeks of debugging memory
allocation problems with LwIP. I don't believe that other people
have got proper applications running with LwIP, because if they
had, they would have run into some of the many memory leaks
and allocation inconsistencies that I have discovered.

But my sources are constantly diverging from the main
tree, so it's getting difficult to keep things in sync, or to
submit patches. An suggestions would be welcome.

One must remember that there is a huge difference
between:

- getting a single TCP session to transfer some short
stream.

- getting a complex TCP application to work
indefinately, with no allocation errors, memory
corruption, or memory leaks, that connects to
thousands of possibly very buggy Internet hosts
each day, using many different services.

An idea that seems right and logical quickly
becomes absurd in the face of a core dump.

This is why I wrote PaulOS. I want something
that is robust according to MY definitions. Most
of the supposedly good code out their breaks
when you use it in a real life situation. A PaulOS
router is what I now browse the web with everyday
and for months since. So I KNOW its solid when it
comes to selling routers. I also KNOW that when
people tell me that a piece of code is cool, they
don't know what they are talking about, because
I know of bugs in that code they havn't even
tested enough to see, let alone fix.

Adam has done a brilliant job, and his code
is of the cleanest I have ever seen. But at
some point emperical experience has do
be added to verify the robustness of the code.

The first thing I can say is that a multithreading
LwIP will never work. Is there anyone out
there who has a production multithreaded LwIP
implementation? I bet no one. My guess is
that any multithreaded implementation will
run for a hundred connections and then die.

Most of the people on this list have all these
visions about LwIP. But have they actually
got code working? Have they got rubust
Internet applications running that don't
crash after two days?

When someone can tell me that they are
USING lwIP effectively, then they are in
a position to talk about where lwIP should go.
Before that time, it is all just hot air.

Thanks to you Duncun, I do appreciate your
interest in PaulOS.

> I didn't get back to you the other week about your sockets implementation - 
> got distracted... From what I could see from your patch on the lwip website, 
> your sockets layer is GPL? that precludes me from using it unfortunately... I 
> was looking around the PaulOS web site for source as well (operating systems 
> are something i'm interested in, so i enjoy reading about them), but couldn't 
> find any - have you published it, or am i just blind? 

My sockets layer is GPL, but is not usable because it can only really
be plugged in as a PaulOS driver. So I don't know why you would
want to include it at all. Its just an example to show how it might
be written.

-paul


-------
On Monday 19 August 2002 20:05, you wrote:
>
>
> Has anyone else has audited lwIP with
> two completely seperate memory tracer
> packages and TWO MONTHS of continuous
> loading? My work may go to waist.

I, for one, would be very interested in this. We've made some small 
modifications to the memory allocation, but haven't tested it in anger... 

Hopefully the list will be able to come to some agreement about the running 
of lwip from now on - its well past time. 

Dunk
[This message was sent through the lwip discussion list.]


Powered by World Online - http://www.worldonline.co.za



------=_Part_1191_3913634.1029844245506--

[This message was sent through the lwip discussion list.]