[RSSAC Caucus] [Ext] FOR REVIEW: RSSAC026v2: RSSAC Lexicon

[Brian Dickson]

On Wed, Jan 29, 2020 at 4:47 PM Paul Hoffman <paul.hoffman at icann.org> wrote:

> Wes' formulation seem close to me. I disagree with the proposal that this
> document should define "anycast" and "anycast routing instance" or "anycast
> site": that won't help almost any reader of the document. Those things are
> important to the operators, but there is enough variety of operations that
> nailing those down by definitions would not be helpful.
>

I think the question isn't whether "anycast" is necessary, it is rather
whether including "anycast" (definition and reference) improves
understanding, including to those less familiar or less technical.

Having a fairly high-level/generic description of anycast, along with a few
examples, might suffice, rather than having to nail down definitions.
Examples could include the following:
- External BGP announcements from a stand-alone instance (not hosted on
anyone else's network infrastructure)
- Internal BGP announcements, when instances are hosted on someone else's
network infrastructure (including the possibility of a single external BGP
announcement effectively covering multiple instances learned via internal
BGP)
- Non-BGP announcements, when instances are hosted on someone else's
network infrastructure. Those could include OSPF or ISIS.

The high-level description of an anycast instance might need to indicate
that what distinguishes a single instance from something that contains
multiple instances, is whether all the hosts that can be reached on the IP
address of the RSI, are behind a single router or a small number of routers
(e.g. in an ECMP scheme).

It may be sufficient to explain that from a particular vantage point, it is
expected that all queries in a small time window, should only reach a
single anycast instance, although they may reach a variety of servers
within that instance.
Two distinct vantage points will either hit the same instance, or two
different instances.
It may also be helpful in describing the nature of anycast instances
creating what amounts to a "watershed" as far as vantage points are
concerned.

I.e. what the definitions should try to do, IMHO, is allow a person
unfamiliar with anycast to understand why they see what they see when
looking at root server identities, regardless of whether the word "anycast"
is used.

Brian
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mm.icann.org/pipermail/rssac-caucus/attachments/20200130/61119fdf/attachment.html>