NFD » NDN Specifications

Please submit revision to the spec on gerrit (http://gerrit.named-data.net/#/admin/projects/NDN-TLV).

I see no problems with the spec, so we can easily merge it. The only question is which branch we will merge it, but we can figure it out later.

We already have this Redmine issue. If we use Gerrit, then comments on the change will disappear down the Gerrit hole. Can we keep this discussion on the proposed changes here in Redmine and skip Gerrit?

Or, can we enforce the discipline that if there is a Redmine issue, then you can use Gerrit to look at code diffs, but comments must be made in Redmine? The problem with Gerrit is that there is the temptation to click and make a line comment in Gerrit which only a few people ever see. We already have a Redmine issue, so comments should go in Redmine.

I pushed a patch set to change the signature value from 2 to 3. See:
http://gerrit.named-data.net/#/c/890/

There are at least 2 major problems with using Gerrit for documentation changes:

1. Many people who are interested in the changes to the document don't know how to read Sphinx code, so the diffs in Gerrit won't make any sense. That's why we should use Word with revision marks, or other formats that everyone can participate in.
2. Someone had made a line comment on the first patch. But it is not visible in this Redmine issue, and because there is a new patch it is not visible when someone goes to the Gerrit page. This is what I mean be the "Gerrit hole" for comments and why we should not use Gerrit to comment on document changes since the history is lost.

The review process should be on gerrit. The history is never lost there. All comments are visible: "Comments" section at the bottom of gerrit issue shows how many comments on which patch set were made.

Our spec is in sphinx format, which is basically plain text file with several formatting conventions. The important part are not the conventions, but text. So, doing it in word just complicates things, as one need to look into many places and then the changes should be converted back to sphinx.

People who are interested in tracking review changes should subscribe to gerrit updates, the same way they subscribed to redmine updates. There is no differences here.

Hi Alex,

You say there is no difference between reviewing chances on Redmine and Gerritt, but there is a crucial difference which makes Gerritt too difficult to use. If someone gets a link to a Redmine issue, they can easily review all the discussion, as they can on this page. But if someone gets a link to this:

http://gerrit.named-data.net/#/c/907/

then they have to dig and dig to review the discussion. For example, in that link, someone started a line comment in patch 1 about a high-level issue of what fields should be in a KeyLocator. Now there is a patch 2. That discussion is buried in patch 1. And even if you open the patch, sort through all the diffs, and find the line comments, then the initial screen only shows the beginning of the sentence. You have to dig again to open the comment to see all the text. Again: Gerritt is a very poor way to hold an online discussion. It is unreasonable to expect that someone who gets a link like this

http://gerrit.named-data.net/#/c/907/

will take the time to reconstruct the thread of a discussion as it spans multiple patches and is intertwined with other unrelated minor comments.

Hi Jeff,

I think Redmine and Gerrit serve different purposes. Redmine is used to discuss the general idea, while Gerrit is used to review the detail of implementation or documentation. We should use both of them.

I think this is a discussion that deserves its own redmine issue. :)
(Seriously - this is a good process related discussion but perhaps needs to be split out from the technical issue at hand.)