Re: port to the BeBox?

["Roy T. Fielding" <fi...@avron.ICS.UCI.EDU>]

> Anyone want to hazard a guess as to the difficulty of a port to the BeBox?
> Greater than the OS/2 port?  Greater than a Windows NT port would be?  
> Would it make sense?

I got a demo of the BeBox when I was up in the Bay Area (they are in the same
building as EIT).  It would not make sense to do a straight port of Apache,
since the BeBox's primary feature is the underlying database instead
of a filesystem.  Thus, you would want an httpd which takes advantage
of it for speed, multithreading, memory caching, and event-based updates.

However, it looked to be about six months away from third-party software
development, except for those who desperately need to be on the bleeding
edge (or want one heck of a MIDI box).  Actually, I think most of its
abilities would be wasted as an httpd server -- perhaps they should build
a cheap BeServer that ups the networking and removes the multimedia.

....Roy

p.s. Yeah, I know this thread is old, but I'm just now getting through
     mail that came in while I had the flu.

Re: port to the BeBox?

[Nathan Schrenk <ns...@neog.com>]

On Wed, 31 Jan 1996, Roy T. Fielding wrote:

> > Anyone want to hazard a guess as to the difficulty of a port to the BeBox?
> > Greater than the OS/2 port?  Greater than a Windows NT port would be?  
> > Would it make sense?

A port to the BeBox might be a bit of a pain, but I doubt it would be 
more difficult than a WinNT port.  The latest Be newsletter has an 
article about the BeOS's TCP/IP networking capabilities.  TCP/IP support 
is not built into the kernel, it is implemented as a 'server'.  The 
article says that there are 2 main problems in porting Unix networking 
code to the BeBox:

1) sockets aren't normal file descriptors under the BeOS, so you can't use 
read(), write(), close(), etc on them.  You have to use send(), recv(), 
and closesocket() on BeOS sockets.  I don't see this as being too big of a 
problem with porting Apache.

2) sockets are not inherited across fork() calls.  I see this as being a 
bigger problem.


> I got a demo of the BeBox when I was up in the Bay Area (they are in the same
> building as EIT).  It would not make sense to do a straight port of Apache,
> since the BeBox's primary feature is the underlying database instead
> of a filesystem.  Thus, you would want an httpd which takes advantage
> of it for speed, multithreading, memory caching, and event-based updates.

I was not aware that the BeOS filesystem was a database.  In the things 
that I've seen describing the BeBox's storage, each device (hard disk 
partition, floppy, etc) has a volume consisting of a database and a 
normal filesystem.  So not all files are stored in the database.

I think that the primary feature of the BeBox is that it is a dual 
processor, high performance machine with integrated MIDI, IR ports, 
and lots of other IO, running a whacked-out new OS.  :)

> However, it looked to be about six months away from third-party software
> development, except for those who desperately need to be on the bleeding
> edge (or want one heck of a MIDI box).  Actually, I think most of its
> abilities would be wasted as an httpd server -- perhaps they should build
> a cheap BeServer that ups the networking and removes the multimedia.

I should be getting a BeBox in the mail next week.  I'll let any 
interested parties know what I think about it.

Nathan

> ....Roy

--
Nathan Schrenk						nschrenk@neog.com
Neoglyphics Media Corp.                              http://www.neog.com/