[openbeosnetteam] weekly news report
- From: <j-schwerer@xxxxxxx>
- To: openbeosnetteam@xxxxxxxxxxxxx
- Date: 7 Feb 2002 01:12:41 MST
Hi all,
Granted the news report has not really been a weekly kind of email yet bu=
t it
should have been so and it is going to be sent more regularly: I am now
keeping track of the OBOS network team things to do and deadlines thanks =
to
postgresql's port to BeOS [should have said thanks to David :-)...]
- Document named "general" (from CVS)
>we should use a version of the mbuff's that the BSD stack uses with chan=
ges
>so we use them more efficiently for our stack.
Don't really know much about those mbuffs.... Is that what you are curren=
tly
working on and committed to the CVS? Could you please tell us a little mo=
re
about that?
>Userland vs Kernel Land
Well from the latest discussions on this list I guess this would have to =
be
changed ^_^. We can do a "one-copy" thing: make a copy when switching fro=
m one
land to the other and otherwise always use pointers and references.
=46rom what I understood the reason why Be's net_server was so bad with r=
egards
to this copy thing is because they *always* copied the data, everywhere (=
from
driver to ethernet, from ethernet to IP, from IP to TCP and so on).
>Initial Protocols
I believe PPP should be part of them and not an extra, shouldn't it?
>Structure if_frame
What if you have more than 3 layers? e.g. with IPv4 encapsulated over IPv=
6
transported over PPPoA....? That would make a lot of extra layers without=
entries in the structure :o)
- "Milestones" (CVS doc)
About the IPv4 checksum + routing milestone: I'd really appreciate if rou=
ting
were implemented through a module mechanism, just like protocols, so it w=
ould
make it possible to support RIP of course, but also at a later time some =
other
mechanisms (OSPF and some Cisco-like i-BGP flavours come to mind).
- "Catching up with old stuff"
from IRC log:
>10:46 < BGA> BTW, who is the project lead? I want to know
>who should I report to. :P
(...)
>10:47 < frey> Jean Schwerer is the official team lead
>10:47 < BGA> Where is he? :P
>10:47 < frey> He's fairly busy/unresponsive :/
Without Internet connection at work would have been a more appropriate an=
swer
when that was written =3DoP Busy is always true however ;-)
I am conscious I have not been very responsive in november and in january=
but
things are getting better I think (and I hope so =3DP).
>10:47 < BGA> Heh... Then let's elect David the new leader. :)
If David or Emmanuel has enough time to help me with this I'd be glad one=
of
them joins me :)
>10:48 < frey> Jean said he has done some work starting from the nettle
>source. I'm not sure where he is,
>and he is painfully slow with email.
Mea culpa. I believe I'm better at answering email faster now. My answeri=
ng
time for anything related to OBOS network team should fall below 36 hours=
now.
That said, I would appreciate if you could send me in private any concern=
s,
flames, arguments or bashes you have against me. I would also like if you=
could please send me advice if you think you have some, if you could tell=
me
what you expect me to do that I am not doing now, etc. You can be as vitr=
iolic
as you want to be, I am not a revenge-thirsty kind of guy ;o) You don't e=
ven
need to be constructive if you don't want to :-)
I won't fill a "black list" of people not to talk to, I won't flame back,=
and
I won't share what you write with anyone else unless you request me to do=
so.
So please, go, shoot, flame, burn, explode, kill! :o)
>10:49 < zathras> what's nettle?
A library that implements a C++ API over traditional sockets. Sounds fami=
liar?
Well, it is:
the BeOS network API *is* built on top of Nettle. However, having dug int=
o
Nettle I can tell 2 things:
- I hate the API it implements, for it doesn't add any simplicity, it's j=
ust
"C++ sockets"
- The code inside it looks rather ugly (from my humble point of view). On=
e
thing that would tend to confirm this feeling is that some people experie=
nced
problems when using the BeOS network API with BONE betas (server connecti=
ons
wouldn't listen) while when those same people were using traditional list=
ening
sockets their programs would work.
Additionally I don't understand why Be built their API on top of Nettle
instead of reusing Nettle code in the API. Nettle is not big or useful en=
ough
to justify being kept as separate classes and still exported by the BeOS =
API
library.
Have to go to work, more later ^_^
Jean
____________________________________________________________________
Get free e-mail and a permanent address at http://www.amexmail.com/?A=3D1=
Other related posts:
