On 03/22/2010 11:14 AM, Mark Handley wrote:
> On 22 March 2010 16:46, Ben Greear<[email protected]> wrote:
>> I don't actually know who has the authority to discuss bigger changes
>> (like what is found in my xorp.ct tree). I want to come to some sort
>> of understanding with them before I put any real effort into breaking
>> out small patches for upstream SVN.
>
> I guess the buck stops with me, but Atanu and Bruce have a better
> overview of the state of all of the code at this point than I do. So,
> in effect it's the three of us right now.
>
> We're working out new processes at the moment, as things settle down.
> The idea is certainly to be much more open to new committers than
> Xorp, Inc was. It's essential we use the energy in the Xorp community
> to best effect, something Xorp, Inc never figured out how to do.
>
> With regards to the xorp.ct tree, could you summarize the main
> differences between your tree and the sourceforge tree for the list.
> I'd like to give everyone a chance to comment on whether there are any
> issues with regards to merging the two trees.
Well, it's been 2-3 years since I started...but the basic ideas are below:
* Allow OSPF, RIP, BGP, OSLR, and PIM to better handle interfaces being
added to and deleted from a running xorp instance via the 'xorpsh' CLI tool.
This involved removing a lot of asserts and assumptions that
interfaces existed in the kernel at a specific time.
For instance, there is no reason to assert if you are trying to remove
an interface that doesn't exist.
Other changes in FEA allow users to configure interfaces that don't
yet exist..this 'configured' state is
remembered in the case that interfaces later appear. I wrote this
several years ago, so it's a bit fuzzy.
Also, Pavlin committed some of these changes..but I think most of them
are still in just my tree.
I also optimized code to remove some 1-sec sleeps on 'commit' so that
driving xorp from xorpsh
programatically doesn't take nearly so long (about .1sec for most OSPF
related commits that we
do in our mobile network emulator).
* PIM: Fix state-machine lockup due to failure to advance state machine in
some error cases.
* RIP, BGP, PIM: Allow binding the protocol sockets to specific interfaces.
* FEA: Support binding to specific interfaces, including multicast
protocols. Use Netlink (Linux only) to filter events so that a single instance
of xorp
pays attention to only it's interfaces.
* FEA: Support experimental Linux kernel API to provide multiple multicast
routing tables per kernel instance. This requires a patch to Linux as well.
Please contact [email protected] if you want to try using this. Xorp.ct
is backwards compatible with normal kernels, so unless you specifically want
this
feature, you do not need any additional Linux patches.
NOTE on mcast: I'm working with a kernel developer to get these changes
into the linux kernel, probably 2.6.35. The API might change
a bit. The current code in xorp.ct is backwards compat, so it could
go in anyway and then I can re-work it proper
when the linux API is finalized.
* OLSR: Enable building OLSR (it is broken in upstream SVN), support
binding to a specific interface.
To build with OLSR: scons enable_olsr=yes
* Support building IPv6 multicast support on Linux. Haven't tested
functionality yet.
Lately, I've added a large amount of relatively boring & repetitive changes to
allow compiling
out much of the IPv6 code in order to allow reduced footprint compiles for
smaller systems.
Also, enable building without boost. I know Bruce likes boost, but I'm not a
big fan of
the extra dependencies. Either way, my current changes allow one to compile
with boost if
they prefer. If someone ever puts in a large amount of boost code that I can't
easily work
around, then perhaps we can remove the option to NOT depend on boost.
I can post a full diff, but it would be so huge it would not really be
reviewable by
any mortal! I can also break out individual diffs, but I have several years
worth, and
I doubt anyone has time or interest to review them all in detail either.
Thanks,
Ben
>
> - Mark
--
Ben Greear <[email protected]>
Candela Technologies Inc http://www.candelatech.com
_______________________________________________
Xorp-hackers mailing list
[email protected]
http://mailman.ICSI.Berkeley.EDU/mailman/listinfo/xorp-hackers
