Change Details

As part of T334238 we want to move traffic from the parsoid endpoints to core endpoints. This would be a lot easier if we didn't need to implement active purging for the new URLs. It looks like we don't have to, and we can get away with a low TTL (a couple of minutes) instead. THis is based on the observation that the cache hit rate for endpoints under /api/rest_v1 is extremely low. This would allow us to greatly reduce the complexity of the system architecture. We can turn off storage and pre-generation of Parsoid in RESTBase (pending T330036), and we can stop emitting purge events from RESTBase. RESTBase can be come a simple procy for Parsoid. Perhaps we don't even have to make any changes to Parsoid in RESTBase, we just re-route the /api/rest_v1/page/html and friends directly to the parsoid endpoints in MW.

As part of T334238 we want to move traffic from the parsoid endpoints to core endpoints. This would be a lot easier if we didn't need to implement active purging for the new URLs. It looks like we don't have to, and we can get away with a low TTL (a couple of minutes) instead. This is based on the observation that the cache hit rate for endpoints under /api/rest_v1 is extremely low. This would allow us to greatly reduce the complexity of the system architecture. We can turn off storage and pre-generation of Parsoid in RESTBase (pending T330036), and we can stop emitting purge events from RESTBase. RESTBase can be come a simple procy for Parsoid. Perhaps we don't even have to make any changes to Parsoid in RESTBase, we just re-route the /api/rest_v1/page/html and friends directly to the parsoid endpoints in MW.

As part of T334238 we want to move traffic from the parsoid endpoints to core endpoints. This would be a lot easier if we didn't need to implement active purging for the new URLs. It looks like we don't have to, and we can get away with a low TTL (a couple of minutes) instead. THisThis is based on the observation that the cache hit rate for endpoints under /api/rest_v1 is extremely low. This would allow us to greatly reduce the complexity of the system architecture. We can turn off storage and pre-generation of Parsoid in RESTBase (pending T330036), and we can stop emitting purge events from RESTBase. RESTBase can be come a simple procy for Parsoid. Perhaps we don't even have to make any changes to Parsoid in RESTBase, we just re-route the /api/rest_v1/page/html and friends directly to the parsoid endpoints in MW.