Re: [homenet] Implementations of IS-IS [was: Routing protocol comparison document]

[Margaret Wasserman <[e-mail]>]

Thanks, Juliusz.  This was very helpful to me in understanding the status of these implementations.

Margaret

On Feb 16, 2015, at 10:58 AM, Juliusz Chroboczek <[e-mail]> wrote:

>> You keep mentioning that Quagga is "alpha quality" but i am not sure that
>> provides any pertinent information.
> 
> Oh, sorry if I wasn't clear.  Quagga has excellent, production quality
> implementations of OSPF, RIPv2 and BGP, and a simplistic but workable
> implementation of RIPng.  On the other hand, their IS-IS documentation
> says:
> 
>    isisd is ALPHA quality at the moment and hasn't any way ready for  pro-
>    duction use.
> 
>    isisd eats bugs for breakfast. If you have food for the maintainers try
>    http://bugzilla.quagga.net
> 
>> If I understood correctly, you mentioned in a private thread that the
>> Quagga IS-IS codebase is not being maintained right now.  Is that correct?
> 
> According to their commit history, the last significant changes date from
> November 2012.  Of course, it is possible that work is being done out of
> tree, perhaps David can clarify.
> 
>> When you add features, such as source specific routing, to your Babel
>> codebase, do you go through a full formal release process with Alpha, Beta
>> and Production versions?
> 
> Significant new features live in a separate branch, usually for months or
> even years.  They are carefully reviewed, then merged into trunk.  After
> they've had some time to (not) break, I make a release.
> 
> For example, the RTT-based metric was implemented by Baptiste Jonglez
> during the summer of 2013.  It was exhaustively tested, and almost
> completely rewritten in November 2013.  It was carefully reviewed by
> myself in April-May 2014, and merged into trunk on 16 May 2014.  It was
> further tweaked by both Baptiste and myself, all the tests were redone,
> and it was released as babeld 1.5.0 on 22 May 2014.  So that took almost
> a year, notwithstanding the friendly bullying by the company who wanted
> the feature (no, they don't pay me, and I'm not under NDA; full disclaimer:
> they bought me lunch once).
> 
> Similarly, the source-specific extension was first implemented by Matthieu
> Boutier in the spring of 2013, then extensively modified.  We have been
> doing serious testing since early 2014.  I've started systematically
> reviewing his patches in the summer of 2014 (which involved some changes
> to the data structures, Matthieu wasn't pleased), but then I ran out of
> time.  I've set aside some time in February to finish the merge, at which
> point I plan to let it simmer for a few weeks before releasing 1.6.0.
> 
>> You have mentioned that Babel is "in OpenWRT" or something along those
>> lines… Is the version with source-specific routing included in OpenWRT
>> yet?
> 
> Yes, in the package babels, which is separate from the mainline babeld
> package.  Users have been complaining about this situation, which is why
> I need to finish the merge as soon as I have the necessary peace of mind.
> 
>> what is the process by which that will be included in an OpenWRT release?
> 
> After I release babeld-1.6.0, hopefully in early March, Gabriel Kerneis
> will include it in OpenWRT (he usually does so within a couple of days)
> and Stéphane Glondu will include it in Debian (he usually takes longer).
> At that point, we'll discuss on babel-users whether we want to delete the
> OpenWRT babels package, or make a dummy package to ease the transition,
> and Gabriel will make the final decision.
> 
>> How many commercial router vendors have it enabled in their products?
> 
> None, as far as I'm aware.  It is widely available for researchers to
> experiment with, and is enabled by default in CeroWRT.
> 
> I hope that this clarifies things, please do not hesitate to tell me if
> you feel that I make claims that are not backed by hard data.
> 
> -- Juliusz