Martin,

On Mon, Nov 10, 2014 at 08:34:23AM +0100, Martin Bjorklund wrote:
> Jeffrey Haas <[email protected]> wrote:
> > One item I would like to proactively draw the working group's attention to
> > is the mount-point drafts:
> 
> Jeff, can you elaborate?  I don't understand how the mount technique
> is related to I2RS ephemeral state.  mount is a way to incorporate
> data from some datastore into another, and this seems to be orthogonal
> to I2RS requirements for ephemeral state.

The main reason to draw I2RS attention to the draft is not that it intends
to solve the ephemeral state issue but that the remote-mount technique runs
into the same consistentcy issues that were problematic for ephemeral state
systems when the ephemeral state had become out of sync with the local
static config.

The remote-mount draft suggests an "eventual consistency" model.  While that
concept is not directly analogous to our likely solution space, the need of
the system to be able to deal with temporary inconsistency is one that has
good use when talking about our ephemeral state problems.

A second thing that I found useful about the draft is that, should I2RS
decide to not permit overriding of configuration state via ephemeral state,
the mount-point concept provides some use cases to demonstrate what you
could do.  The ability to override hasn't been explicitly accepted as
something we *must* do in I2RS and has been a source of something that makes
things hard.

-- Jeff

_______________________________________________
i2rs mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/i2rs

Reply via email to