Gary, that sounds fine though I haven't looked into the precise details.
Martin, the wadl urls are static but content changes daily; etag is probably much better than a very short cache halflife for clients.
Some thoughts:
- we need this to be robust on rollouts (as icing is)
- I don't care if its in a new dir or not
- we release new API versions very rarely so having a manual step there is fine (but lets capture the know-how somewhere!)
- writing /anything/ to well known files on disk for testing is a step backwards in the parallel testing efforts (as would requiring apache during test runs) - please avoid that (e.g. by retaining the serve-from-zope facilities we have today for development environments).
Gary, that sounds fine though I haven't looked into the precise details.
Martin, the wadl urls are static but content changes daily; etag is probably much better than a very short cache halflife for clients.
Some thoughts:
- we need this to be robust on rollouts (as icing is)
- I don't care if its in a new dir or not
- we release new API versions very rarely so having a manual step there is fine (but lets capture the know-how somewhere!)
- writing /anything/ to well known files on disk for testing is a step backwards in the parallel testing efforts (as would requiring apache during test runs) - please avoid that (e.g. by retaining the serve-from-zope facilities we have today for development environments).