On Wed, 15 Sep 2010, Andy Newton wrote:
Forgive me if this is a bit naive, but I was under the impression
that the use of rsync gave us this type of capability for free. It
is not evident to me what benefit this protocol provides unless
there is some desire to mesh together structures.
This is about consolidating publication points and (for some) avoiding
running an rsync server. It also reduces the number of rsync sessions
needed to retrieve the whole PRKI. Meshing structures, by which I
assume you mean putting objects in a common directory structure, is
necessary to the end, but isn't so much a goal in and of itself.
Additionally, from the operations considerations in Section 4:
Placeholder section to talk about nesting children under parents in
the sameso repository, to allow for a single rsync to fetch both
(observing that the rsync setup times tends to dominate over the sync
time). And, more distressingly, talk about the access control
impacts of that nesting.
Why isn't the rsync observation a general concern? Or is this
statement an indication that this protocol is needed to get around a
problem with rsync? If the latter is true, wouldn't it be better to
address the cause of the problem rather than to route around it?
rsync setup times are not a general concern because we're expecting to
reduce the number of rsync sessions a relying party much use to fetch
the whole RPKI down to something manageable. Consolidating
publication points serves several purposes and would probably be
desirable even if the rsync problem were fixed.
From an engineering standpoint, I think Rob, et. al. have taken the
view that rsync is a take-it-or-leave it thing, slow setup included,
and moving to an entirely different transport isn't worth the effort.
-- Sam
_______________________________________________
sidr mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/sidr