Request: Ability to Add Headers to Log

[prabhakm]

Hi folks, John Riviello(github account: @JohnRiv) and I (Prabha, github: @prabhakm) have been following the NEL spec and are looking forward to being able to use it in browsers soon. One thing we would like to see added to the spec is the ability to include the values for headers we specify in advance that we are interested in.

In particular, we would like this because we make use of distributed tracing, and all of the major options in that space including but not limited to Zipkin, Jaeger, OpenCensus, Opentracing, etc -- make use of headers to pass tracing data along. If we could somehow include those tracing headers in the NEL report, then we would be able to directly associate a Network Error with a specific trace to be able to understand what a user experienced.

Any thoughts about this? Thank you.

[dcreager]

Up to know we've only talked about how to add additional server-generated metadata — and for that, we've suggested (for instance, in w3c/reporting#61) that the best way to do that is to add some extra URL parameters to your collector URLs. That would allow you to pass opaque data from the origin server, through the user agent, to the collector.

The tracing information is an interesting example, since the origin server can't know in advance which trace ID the user agent will create for the request, so the URL parameter technique won't work. We'd need something like you suggest — your NEL policy could include a list of headers to include in NEL reports, and the report payload would copy the values of those headers from the request.

My main worry is whether we'd need a list of unsafe headers that you could never add to your NEL policy — or conversely, a list of safe ones that are the only ones you're allowed to include. All of the request headers would be visible to server if the request succeeds, so this might not be necessary.

[JohnRiv]

Thanks @dcreager. I noticed @AloisReitbauer from the W3C Distributed Trace Context Working Group expressed interest in figuring out how Server Timing and Distributed Tracing can work together... perhaps there's an opportunity to make Networking Error Logging and the Reporting API part of that discussion as well.

[dcreager]

Thanks for the link, John! I agree that this could be a great place for us all to link up; I’ve followed up on that issue.