[openbeosnetteam] Re: Some Questions...

> I am now starting to work on some code - Couldnt let David and Axel have
all
> the fun!  I am actually studying my Masters externally aswell as working
> fulltime, however, I have just dropped one of my units so I could put some
> time into this...  How is that for proiritizing!!!

Excellent!

>
> I have some questions to start with...
>
> 1.     The major question, I have never used CVS to check in before, only
to
> check out... I am not actually sure how to check in... Can someone please
> give me a breif run down - if not Im sure I can find out the information
> somewhere...

OK, ask specific questions and we'll be able to help. Otherwise look on the
openbeos web site for an article that gives mroe details. NB, if you haven't
already, get an SF account and email the details to M Phipps so he can grant
you commit priv's.

> 2.    I started to do some work on the ICMP and I am filling in some
> defines/enums.  Are there any header files that we have to conform to in
> this respect, in terms of the names of defines and structures - I have
never
> wirtten code for BeOS before.  Also, I see that in a lot of places 'enums'
> (without labels) are used in place of defines, ie , "enum { ... };".  I
have
> never used enums like this before, what is the benefit of doing it this
way
> in place of a define?  And what and when should we be using them in the
> code?

Examples?

> 3.    What have people planned to do for IPv6 support?  If ICMP4 and ICMP6
> are similar, should we build them in together, or keep them seperate as
IPv4
> and IPv6 are?

IPv6 will be implented as seperate modules. ipv6, icmp6, nd6

> 4.    Does anyone have any thoughts on where the INADDR_LOOPBACK should be
> checked?  I think it should be checked at device level (Ethernet.c)
instead
> of Protocol level.  Any objections?

Checked in what way? Bear in mind we're modular, so everything to do with
ipv4 should really be in the ipv4 module! Also, bear in mind how this should
work...

datagram from loopback -> injected into stream as if from a remote host
datagram to loopback -> we lookup the loopback address and get a pointer to
the ifnet structure concerned with the loopback device driver, and so we
simply call ifnet->output and it goes straight to the loopback_output and
hence straight into the input queue for loopback.

Does that make sense? So, I'm afraid I would have to -1 adding checks at the
ethernet level.

Keep the ideas coming though.  I'll try to commit what I have in the
afternoon but it doesn't currently work after all the changes I've made and
there are a LOT of changes involved!

david

Other related posts: