Moved the canonicalization out of the top level

[iherman]

Did some editing to answer to the issues raised in #349. All changes are purely editorial.

Fix #349.

---

Preview | Diff

[mattgarrish]

A couple of comments:

• it still doesn't seem intuitive to introduce the steps for generating the canonical manifest before any mention is made of it in the specification; that's what threw me the first time. It would be better to have it follow the section on obtaining the manifest, like the section on processing does, than try to link forward to where it matters.
• slightly unrelated to this issue, but why is a json object parsed from text in step 7, but the canonical manifest is created from text again in step 9? Similarly, it looks like creating a canonical manifest results in a json object that could be used for processing, but that step also begins with parsing text. After obtaining the json in step 7, aren't you continuing to manipulate that object throughout the subsequent steps?

[mattgarrish]

Also, rather than calling the section "Canonical Manifest" it might be helpful to provide a name that matches the lifecycle nature of the section, so something like "Generating a Canonical Manifest".

[iherman]

You are right on all accounts:-)

I have changed the order as you suggested, changed the title, and also modified the algorithm. Parsing is only done once, and the result of parsing (json) is input to the canonicalization step. There is also an additional remark in that algorithm that the input is a json object representing the manifest.

Thx:-)

[mattgarrish]

The processing section looks like it still needs some tweaks to accept a json object that is the canonical manifest rather than parsing text, but otherwise looks better now.

[iherman]

I will do that tomorrow morning and then merge. ok?

[mattgarrish]

sgtm