Update Setter used for Injection

[carlosalberto]

From this comment, talking about api-propagators.md:

- the `Setter` invoked for each propagation key to add or remove.

For HTTP, does Setter do header.Set or header.Add?

The W3C tracestate header may be split into multiple entries (repeated headers). It's impossible to do it if Setter only has a Set method.

(But maybe it's ok, since not all transports where we might use HTTPPropagator actually allow repeated headers).

[andrewhsu]

Having a look at this with @carlosalberto and it looks like this issue may be desirable before the trace api freeze because it could change how Extractions Getter interacts.

[tedsuo]

Repeating my comment from #433.

Given that headers can be converted between their single line and multiline equivalents we could require that Set and Get must both use single line representation. Under the hood, the carrier can split and concatenate multiline headers as needed.

A recipient MAY combine multiple header fields with the same field name into one "field-name: field-value" pair, without changing the semantics of the message, by appending each subsequent field value to the combined field value in order, separated by a comma.

The exception is Set-Cookie, but I don't believe we should ever be touching that.

[Oberon00]

Given that headers can be converted between their single line and multiline equivalents

Only if we assume only HTTP will ever be used by the TextMapPropagator
EDIT: Although most other protocols won't support having the same header multiple times, I guess.

[carlosalberto]

Given that headers can be converted between their single line and multiline equivalents we could require that Set and Get must both use single line representation.

Would love that myself. If we can agree on that, I'd be super happy.

[carlosalberto]

Similarly to #433 I think we should state specifically that Setter does a SET operation, rather than an ADD. Later, if truly needed, we can add an Update method.

cc @yurishkuro (original author of the comment ;) )

[andrewhsu]

from the spec mtg today, @carlosalberto will add note about details

[carlosalberto]

Added #976 to clarify this part of the Specification - the actual addition of an Add operation will be considered after GA.