Release Notes

These are not all encompassing, but we will try and capture noteable differences here.

3.1

expose a send_chunk method which will return the bare http response, but will still take care of some of the messier header negotiation
support for alternative names for optional send headers
removed mex-MessageType as a send header, it's not required
removed mex-From as sender header, it's not required
support for setting compress for an individual message (rather than just using the 'transparent_compress' init arg)

Moved to v2 mesh content types, sending application/vnd.mesh.v2+json for all requests, and receiving v2 response types.
Allows server to negotiate compression based on accept header.
For chunked messages, all chunks except the last chunk, should be greater than 5MB in size, though this is not enforced in the client or mesh sandbox to allow testing.
All headers will be lowercase.
Inbox pagination and filtering list_messages supports max_results and workflow_filter passthrough.
new method iterate_message_ids, supporting workflow_filter
new method iterate_messages, supporting workflow_filter
message.status are also lower case accepted, acknowledged, undeliverable rather than Accepted, Acknowledged.

previously send_chunk retries could be optionally enabled by setting the max_chunk_retries when creating the client object this has been replaced with the MeshRetry based on the standard urlib3.Retry object in the HTTPAdapter
By default, requests will have two retries with an exponential backoff of 500ms, but initial send_message requests will not be retried.
If you wish to disable retries completely set max_retries=0
Timings and specific behaviour can be tuned using retry_backoff_factor, retry_status_force_list and retry_methods or completely overriden by passing in max_retries=Retry(...)
NOTE: if retries are enabled, HTTPErrors will be wrapped in request.exceptions.RetryError, ensure catching RetryError as per HTTPError

HSCN endpoints are 'deprecated', and internet facing endpoints should be used in preference. For clarity HSCN endpoints have been renamed, prefixed DEPRECATED_HSCN_.
Primary endpoints are now LIVE_ENDPOINT, INT_ENDPOINT, DEP_ENDPOINT.
TRAIN, OPENTEST, DEV endpoints have been removed, but will still work with the client if you have the url and certificate.

get_tracking_info and track_by_message_id endpoints removed, in favour track_message as v2 api tracking response is significantly different.

MockMeshApplication server removed completely, remaining unique pytests moved from sandbox to pytest-httpserver.