API updates

[William Reade]

On Thu, May 30, 2013 at 2:29 AM, Tim Penhey <tim.penhey at canonical.com>wrote:

> I have to admit that I agree with Ian here.  What is the cost of us
> implementing this?  I think that should be compared with the cost of
> not implementing it.  Particularly in relation to point 5.
>

Yeah, point 5 is good and well made. Strawman for consideration: what if we
just *never* update local fields? Most of the discussion thus far has
centred around never-happen use cases: stuff like Unit.AssignToMachine,
which should surely *never* go in the API, and should thus probably not be
given too much weight...

I'm starting to develop a feeling that this whole point is moot; I am aware
of only one case in which a local object will want to re-read its own data
without necessarily having refreshed is in the unit agent -- in which we
set the unit's addresses on startup, and subsequently read them when
joining relations -- and this is a situation commonly agreed to be crack
for several reasons (apart from anything else, *machines* will soon be the
ones with addresses). Can anyone point me to any other API-relevant case
where object updates are useful or necessary?


> Is it reasonable to have the API caller specify if they want a full
> representation back?
>

That might be interesting, if we can find a use case in the first place :).
Any suggestions?

Cheers
William
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/juju-dev/attachments/20130530/f0a2c64d/attachment-0001.html>