blob: 0da35533675baf6a99e92e5b6415ee3228e1e2c9 [file] [log] [blame]
Quagga TODO list
2013-03-29
This is the Quagga primary TODO list. It is on git because that way changes
pass through the usual process just like code does, therefore they will have
the same visibility.
If you are working on something beyond a simple fix, to avoid double work it
is a good idea to submit a patch to this TODO list when you are starting,
listing what you're doing. Also, as others may have done just that, check
the list before starting.
Google Summer of Code 2013 note: this list double-serves as idea list for the
Summer of Code. Ideas considered suitable for students are marked with a star
after the number, like this: "[Q999*] achieve world peace". They will also
have extended descriptions. Nevertheless, if you'd like to do something else,
just write a mail to the mailing list: quagga-dev@lists.quagga.net
"GSoC-Mentors:" listings are preliminary at this point.
Overall
=======
[Q000] improve unit test architecture
[Q001] kick invalid runtime tests from configure.ac, use list of supported
OSes and their APIs instead.
Priority: low
State: patch half-done 2013-03-29 David Lamparter
[Q002*] clean up zebra IPC, remove code duplication, align to common API
Priority: high
GSoC-Mentors: David Lamparter, Christian Franke
Quagga posesses an IPC mechanism to exchange route information among
the different daemons and Zebra, the kernel-interface. This mechanism
is implemented in libzebra, but is currently used in all sorts of
different ways in the individual protocol daemons. Also, in the future
the entire protocol needs to be redone in an extensible way, so we're
able to support MPLS, BFD, Multi-Topology/Instance, VRFs, ...
This TODO entry only refers to the first-step API cleanup. All the
daemons need to use a single, well-defined libzebra API. Only after
this has been addressed can we look upon changing the protocol itself,
since by then it will be encapsulated inside libzebra.
[Q003] add multi-instance / multi-topology support to the individual protocols
[Q004] MPLS support
State: work in progress 2013-03-29 Renato Westphal, Timo Teräs
[Q005] BFD support
State: two old implementations exist, contact Hasso Tepper
library
=======
[L000] improve route_table speed, eg strided lookups for common prefix depths.
[L001] ipv6 addresses need concept of valid/preferred
[L002] implement a generic daemon access/control protocol (eg D-Bus like?
simplified SNMP-a-like? NETCONF?)
[L003] extend vty command definitions to allow them to be self-documenting
i18n command help strings
[L004] create a common libspf (for ospfd, ospf6d and possibly isisd and more).
cf. TODO item [O000] for the ospfd/ospf6d specific variant
[L005] stabilise the API (possibly including symbol/library versioning voodoo)
[L006] Document the exported API (DocBook/Doxygen?)
[LE00] incorporate library changes from Euro-IX branch, except threading
[LE01] incorporate threading library support from Euro-IX branch
zebra
=====
[Z000] Pointopoint address configuration.
Priority: low
State: patch done & tested 2013-03-29 David Lamparter
[Z001] Add support for valid and preferred lifetimes to IPv6 addresses
[Z002] proper support for (at least) 1-level recursive routes
Priority: high
[Z003] Ability to set src on routes, where systems support it.
[Z004] Ability to apply route-maps to daemon route updates.
bgpd
====
[B000] HUP signal support (reload configuration file).
[B001] BGP multi-path extension, relaxed mode
Priority: medium
[B002] move FSM state to be per-connection, not per-peer.
[B003] Add support for internal and minimum-metric MED setting
ripd
====
[R000] Multipath support.
ospfd/ospf6d
============
[O000] move SPF to common code
[O001] extend code sharing between ospfd and ospf6d beyond SPF
[O002*] OSPF testing replay tool
Priority: medium
GSoC-Mentors: Martin Winter, Christian Franke, David Lamparter
In order to extensively test OSPF implementations, a tool to fake an
OSPF neighbor is immensely useful. This tool needs to be capable of
forming an adjacency and pushing LSAs to the device to be tested. To
maintain the adjacency, some minimal state tracking is useful.
In total, the tool needs to form an adjacency, read and push LSAs, and
output received LSAs. Additional tools to generate LSAs from
specifications as well as verify received LSA correctness can then be
built on top of that.
The tool needs to support IPv4 and IPv6, possibly split into 2 tools
with some code sharing.
ospfd:
[O400] Demand circuits.
Priority: very low
[O401] Multiple instances.
Priority: medium
[O402] HUP signal treatment.
Priority: medium
State: patch on ML needs review 2012-06-04 Mattias Walström
ospf6d:
[O600] fix ospf6d in general
Priority: high
State: patches tickling in from Cumulus Networks 2013-03-29 Dinesh Dutt
isisd
=====
[I000] reassess isisd TODO
[I001*] IS-IS testing replay tool
Priority: medium
GSoC-Mentors: Martin Winter, Christian Franke, David Lamparter
see [O002*].
[I002] Mesh groups (RFC2973)
[I003] Crypto authentication (RFC3567)
vtysh
=====
[V000] untangle readline specific bits
[V001] add a vtyd with a vty (ie telnet) frontend (as opposed to readline)
[V002] (=> [L002]) use daemon control protocol
[V003] better AAA support than just PAM, eg krb5, SASL, LDAP...