Skip to content

fix(ndt7-protocol.md): clarify test "runtime" concept#426

Merged
bassosimone merged 4 commits intomainfrom
bassosimone-patch-1
Feb 27, 2026
Merged

fix(ndt7-protocol.md): clarify test "runtime" concept#426
bassosimone merged 4 commits intomainfrom
bassosimone-patch-1

Conversation

@bassosimone
Copy link
Copy Markdown
Member

@bassosimone bassosimone commented Feb 27, 2026

Existing historical clients (e.g., the minimal JavaScript client and the minimal Go client) did not have a consistent definition of the runtime, with the JavaScript client erring on the side of including the TCP+TLS+WebSocket handshake time and the Go client, instead, erring on the side of not including it.

This diff clarifies the matter. After a conversation with @robertodauria, and taking into account m-lab/ndt7-js#111 and m-lab/msak-js#23, we determined that it is probably better to adopt the behavior originally implemented by the minimal Go client.

The rationale for doing this is fairness. The original reason for killing a test unconditionally after ten seconds in JavaScript was grounded in UX concerns (do not keep the user waiting for more than 20 seconds overall). However, with high RTTs, this also means providing a less useful overall goodput estimate. To remediate this, we allow the client to run for more time.

Refer to the aforementioned pull requests for a more-in-depth analysis of this behavior, which helps reconstructing the chain of thoughts that led us to implement this spec change.

Since this diff is a spec clarification that correctly defines the concept of runtime, I am opting to bump the minor version of the spec.


This change is Reviewable

Existing historical clients (e.g., the minimal JavaScript client and the minimal Go client) did not have a consistent definition of the runtime, with the JavaScript client erring on the side of including the TCP+TLS+WebSocket handshake time and the Go client, instead, erring on the side of not including it.

This diff clarifies the matter. After a conversation with @robertodauria, and taking into account m-lab/ndt7-js#111 and m-lab/msak-js#23, we determined that it is probably better to adopt the behavior originally implemented by the minimal Go client.

The rationale for doing this is fairness. The original reason for killing a test unconditionally after ten seconds in JavaScript was grounded in UX concerns (do not keep the user waiting for more than 20 seconds overall). However, with high RTTs, this also means providing a less useful overall goodput estimate. To remediate this, we allow the client to run for more time.

Refer to the aforementioned pull requests for a more-in-depth analysis of this behavior, which helps reconstructing the chain of thoughts that led us to implement this spec change.

Since this diff is a spec clarification that correctly defines the concept of runtime, I am opting to bump the minor version of the spec.
@bassosimone
Copy link
Copy Markdown
Member Author

Note that coveralls is down due to technical issues.

@bassosimone bassosimone merged commit 1b0cc91 into main Feb 27, 2026
1 of 2 checks passed
@bassosimone bassosimone deleted the bassosimone-patch-1 branch February 27, 2026 15:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants