openstack team mailing list archive

[Thorsten von Eicken <tve@xxxxxxxxxxxxxx>]

On 7/8/2011 5:32 PM, Lorin Hochstein wrote:
> I don't think that the OpenStack project should commit to maintaining 
> EC2 compatibility at all costs, only as long as the benefits outweigh 
> the development costs. In particular, if Amazon deliberately started 
> making changes to break the API, that would be a good time to consider 
> dropping support.
Some observations from the cheap seats:
 - OpenStack will never be 100% compatible with either the EC2 API spec 
or the AWS EC2 implementation. Things like spot pricing, or cluster 
compute instances, or product codes may never make sense, or if that 
type of feature is implemented it may look completely differently. So 
framing the discussion around 100% seems pointless.
 - There seems to exist tremendous value in supporting two APIs in that 
it forces certain architectural structures and certain implementation 
rigors that have great value going forward.
 - Having an EC2 compatible API as an easy on-ramp seems to be quite 
valuable, but I'd be hard-pressed to put numbers behind that. I'd keep 
the goals around the "core" cloud functionality, what you'd expect 90%+ 
of users to need as a minimum to test out OpenStack. I'd expect almost 
anyone to understand that they will need to move to the OS API sooner or 
later. But the compatibility API can make it much easier to get there 
incrementally. Not sure how to draw the line, though.

Thorsten