Re: New directions for the Etherboot project

On May 2, 2005, at 4:36 AM, Geert Stappers wrote:
> On Sun, May 01, 2005 at 11:06:11PM -0400, Marty Connor wrote:
>> Hello Everyone,
>> __Growth and Expandability__
>> Etherboot has grown lot, and is constantly gaining capabilities.  This
>> is a good thing, but after a while a single program can attempt to do
>> too much, and become difficult to maintain and use.
>    [ modular ]
> human / organ / tissue / cell
> even we have smaller parts

Geert, thank you for your comments.

Yes, biological metaphors are compelling.  Mitosis comes to mind.

>   [ announce of  gPXE ]
>> I think we can do this, and more importantly, I think we _should_ do
>> this.
>> What do you think?
> PXE is a poor documented mixture of BIOS and network boot code.
> It is for i80?86 processors, but that architecture is dying.

There is some truth in what you say, but there is are also 
counter-arguments that are easy to make.  I understand your point of 
view, but I am also mindful that technical superiority is not always 
sufficient to support people in an environment where there is a large 
installed base, no Free alternative, and the existing specification, 
while flawed, provides (in multiple implementations) sufficient 
functionality and reliability to have become a usable defacto standard.

We need to evolve in order to grow and remain vital.

> Etherboot is stand alone network boot code.
> It supports several processors.

Well, it sort of does.

The Etherboot code base does not support things by itself -- it is 
developers and testers and users who determine how well it works.  I 
would guess that over 99% or our users are on x86, and of those, most 
are trying to boot a kernel or some version of DOS.

While other processors are interesting, respecting our core user base 
is also very interesting.

Remember, we support _users_, not just _architectures_, and most of our 
users happen to be on x86 hardware.  If someone wishes to support other 
users on other architectures using our code that is a fine thing, but 
unless it is clear that there is an ongoing commitment to support, I 
think we have to drop the code from production releases when 
development support fades.

That may seem like a somewhat hard stance, but evolution is like that.  
Many Linux distributions have to make the same hard choices about which 
architectures to support.  In our case, Sun and PPC already have 
network booting in Open Firmware for the most part.  Embedded targets 
are interesting, but support of them needs to be weighed carefully.

> I do like the "cell approach" and don't like "the PXE mantra".
> But hey, I don't have to work on the DHCP / TFTP mutant called PXE.
> And that is only the network part.

Please help with what you like.  All opinions will be heard.  
Contributions will be judged on both technical merit and how well they 
address needs of our user community.

> At this mailinglist was recently a posting which reported
> some software misses stuff in EtherBoot that is available in properiaty
> PXE. So now we, the EtherBoot project, have a "customer" that wants us
> to find out what BIOS function that software needs.
> The only reason: "It works with other PXE clients"

The only reason _you_ hear.  I hear other reasons, and frankly "It 
works with other PXE clients" seems like a pretty good interoperability 
reason to me.

> Please remember that EtherBoot is about networkboot.
> ( not a  80286 BIOS repair kit )

If you mean the current Etherboot _product_, I can agree to a point.  
If however you mean the Etherboot _project_, that's a different matter. 
  The project is about people, and what we decide together.  It is quite 
conceivable that the Etherboot project will decide that the Etherboot 
product should adopt some PXE capabilities.

It is far too easy to attack PXE and x86 and call names and make fun of 
technology we think is lacking.

It is much harder to create code that works in the real world.

Etherboot has design and implementation flaws as well.  We just don't 
happen to have a specification that people can easily make fun of.  
Beyond this, our documentation is weak, partially because we add 
features but have no specification to update.

So, perhaps we should consider the old saying "People in glass houses 
shouldn't throw stones" :)

> Okay, so far my rant.

Do you feel better now, my friend? :)

> The move from monolithic to modular is good.
> What I think that is needed for that,
> is an overhaul of the build system.

Yes, that's a beginning.

> Geert

I appreciate your comments, and the time you took to write them.

Marty



-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.
Get your fingers limbered up and give it your best shot. 4 great events, 4
opportunities to win big! Highest score wins.NEC IT Guy Games. Play to
win an NEC 61 plasma display. Visit http://www.necitguy.com/?r=20