Re: [sidr] Slides for "RPKI Over BitTorrent" presentation

[Rob Austein <[e-mail]>]

At Wed, 28 Mar 2012 20:33:24 -0400, Danny McPherson wrote:
> 
> i don't think the rsync scale issues surprise anyone that was paying
> attention.  If we're already considering new architectures,
> substrates, et al., here perhaps we shouldn't be so quick on the
> trigger for Standards Track work and move this and related
> "investigation" to the IRTF, or at least ensure they're only
> Experimental until broader experience is gained.

If you're talking about moving all of the existing SIDR protocols to
Experimental, that's a cheap shot and you know it.  Any protocol will
behave badly if misused.  In my opinion the RIRs are currently using
rsync badly, given the way they've chosen to set up their
repositories.  This is not an inherent property of the protocol, just
a bad configuration decision which could be fixed at any time.  It may
be possible to find or invent better protocols than rsync, but rsync
was a reasonable choice for initial deployment.

If you're just talking about RPKI over BitTorrent, the BitTorrent
experiment was just that, an experiment.  The slides say so, and
state, in so many words, that this was not a proposal to change the
SIDR protocol suite.

For the record, this presentation was originally targeted at the IEPG,
as a direct follow-up to multiple suggestions received at a previous
IEPG meeting that it might be interesting to see how BitTorrent works
in this environment.  I presented this at the SIDR WG meeting because
the chairs thought it might be of interest to the WG.

Several other people have made comments which pretty clearly indicate
that they either did not read or did not understand the slides.  If
the former, please do so; if the latter, my apologies, please ask.

> Looking at the charts you presented I can only imagine what will
> happen with 40K RPs and >1M objects (which might be a reasonable
> assumption if this were fully deployed today - and that's only
> focusing on routed number resources).

Thank you for making my point for me.  We can do much better than what
we're seeing right now, but the first step is understanding where the
problems are, which involves studying behavior and reporting results.
This is less likely to happen if every report is viewed as an occasion
for an axe-grinding contest, so perhaps we should focus on the
technical issues.