[editorial] service.md: normalize link to process.md

[chalin]

The link needs to be normalized (to refer to a page, not the page title), otherwise OTel.io link checking fails; for details, see open-telemetry/opentelemetry.io#4253 (review).

/cc @joaopgrassi

[lmolkova]

@chalin thank you!
what's wrong with https://github.com/open-telemetry/semantic-conventions/blob/main/docs/attributes-registry/process.md#process? why it not ok to refer to the title? Is there something we need to change to avoid things like this?

[trisch-me]

@lmolkova it refers to /attributes-registry/process/#process, it's probably a bug for conversion from process.md#process to /process/#process, I think the right way would be to fix instrumentation which does conversion

But the process.md#process doesn't have much sense as well as there is only process now. Maybe it's a leftover from old structure before registry

[lmolkova]

@lmolkova it refers to /attributes-registry/process/#process, it's probably a bug for conversion from process.md#process to /process/#process, I think the right way would be to fix instrumentation which does conversion

thanks for the clarification @trisch-me. I believe this conversion to /attributes-registry/process/#process is not part of semconv tooling and hence we can't fix it or prevent things like this in the future.

[chalin]

... what's wrong with https://github.com/open-telemetry/semantic-conventions/blob/main/docs/attributes-registry/process.md#process? why it not ok to refer to the title?

Because on the OTel.io website, page titles don't have anchors, and hence they can't be linked to. In this repo (and any other OTel spec repo), it is a bit moot to link to the title, since the title is the first thing displayed on the page. So the win-win strategy is to instead just link to the page. Do you agree?

[lmolkova]

@chalin I do agree, the problem is that it's hard to catch inside semconv - process.md#process is a valid link within the repo and does not fail the CI. So we have to rely on the reviewers to catch it, i.e. can't guarantee that new versions of semconv will be compatible with website links.