Re: [babel] draft-ietf-babel-applicability WG Last Call

[Juliusz Chroboczek <[e-mail]>]

Thanks for your comments, Adrian.  I'll give myself a few days to think
them over.

> Suggest to strike the following.

>    Given a sufficiently friendly audience, the principles
>    behind Babel can be explained in 15 minutes, and a full description
>    of the protocol can be done in 52 minutes (one microcentury).

Strongly disagree.  Comprehensibility and implementability are important
properties of Babel (second only to the fact that it works).  I'm open to
rewording if you find the current wording overly whimsical.

> I'm personally a little sceptical about the first example [in Section
> 2.2] because my experience suggests that coders are fiendishly capable
> of constructing bugs that do all manner of unexpected things.

This is based on actual experience.  I'll try to reword it, but I stand by
my claim.

For example, an early version of babeld had a bug that randomly corrupted
buffers; with high probability, the corruption happened within an IPv6
address, so nodes would simply learn spurious host routes that would time
out after a few seconds.  (It took us ages to find the bug, in the
meantime we were running with a network populated with occasional
"martians".)

Another example: I've recently introduced a bug in the sending of
triggered updates (fixed in 1.8.1).  The effect was that convergence was
slowed down in some cases, until the next periodic update, but the
loop-avoidance mechanisms were still functional, so not much harm happened.

I'll think about how to reword it, perhaps summarising the two examples
above might be useful.

-- Juliusz