serialization_id is too long to be used as a score for Redis ordered set

[Anislav]

Activities' serialization_id is used as a score for a Redis ordered set by the timeline storage. Redis scores are represented in IEEE 754 double-precision floating-point format. Converting a serialization_id which is 27 symbols long in this format, results in losing precision. This causes filtering activities to work inappropriately in some occasions.

[tbarbugli]

Hi Anislav; thats right, the default serialization_id implementation is too large to fit in Redis sorted set scores; we are going to add documentation on this and invite users to come up with their own implementation of that.

At this present we are using Feedly ourselves on our own project (getstream.io); there we took a different approach and went for something similar to twitter snowflake (which is pretty much current timestamp + host data + random data).

Unfortunately thats not a good default behaviour either because it breaks most of our users implementations (or at least makes them more complicated).

Tommaso

[TheoRyley]

Is this still the case?