http: support for adding tokenized HTTP headers to local reply mappers

[rulex123]

Commit Message: this PR introduces support for adding tokenized HTTP headers to local reply, which was mentioned as part of #11648.
Additional Description:
Risk Level: low
Testing: new unit and integration tests
Docs Changes: n/a
Release Notes: updated version history

[repokitteh-read-only]

CC @envoyproxy/api-shepherds: Your approval is needed for changes made to api/envoy/.
CC @envoyproxy/api-watchers: FYI only for changes made to api/envoy/.

🐱

Caused by: #13538 was opened by rulex123.

see: more, trace.

[htuch]

I don't fully grok how this works as an end-user. Can you add some docs with a complete example? I.e. how is this any more useful than regular headers_to_add, why doesn't that work with local reply mapper, etc?

[mattklein123]

In addition to the question above: do we actually want to allow using the header formatter? Wouldn't that give us a fully generic way to get what we want?

[mattklein123]

Thanks for working on this! Some API questions to get started.

/wait

[mattklein123]

In addition to the question above: do we actually want to allow using the header formatter? Wouldn't that give us a fully generic way to get what we want?

[rulex123]

@mattklein123 sorry, it took me a while to circle back to this. I have changed the API so that the configuration of a formatter for a header value is supported. Not sure if that's closer to what you had in mind: could you please take a look when you get a chance and lmk what you think? Thanks!

[mattklein123]

Sorry, stepping back, can you describe at a high level what problem we are trying to solve? And then we can figure out the right solution? I'm a little confused what we are trying to achieve (generic formatting, substitution, etc.). Thank you.

/wait-any

[rulex123]

Yep, good idea to take a step back.

The original issue reported here #11648 was that custom headers were not being added to local replies; that was taken care of by this code change #12093.

However there was a larger discussion around the original issue, and a point was made about possibly introducing support for a tokenised header (where each token is a header name/value pair) that could be useful in a scenario where you want to send hop by hop debug info.

I thought that introducing the idea of a tokenised formatter for a header’s value would be one way to achieve this, but I might have misunderstood something or made the wrong assumptions. I hope that clarifies what I am trying to achieve?

[mattklein123]

I thought that introducing the idea of a tokenised formatter for a header’s value would be one way to achieve this, but I might have misunderstood something or made the wrong assumptions. I hope that clarifies what I am trying to achieve?

I think my question is if we support the full substitution formatter (which I think you are saying we do) can't the user just do whatever concatenation they want manually?

/wait-any

[rulex123]

Right, what the heck, the current API and formatters already support all this. Sorry for the confusion, I wasn't thinking straight: closing the PR.