286592 - RDF resource delegate ownership model must be COM-safe

Status: RESOLVED INCOMPLETE

(Core Graveyard :: RDF, defect)

Description

[Benjamin Smedberg]

The current RDF resource delegate ownership model is not COM-safe, and cannot be
implemented in JS safely. I have a fix, and I would like to get it in for 1.8.

Comment 1

[Benjamin Smedberg]

Attachment: Add ownership models

Scott, do I need MOA also?

Comment 2

[Scott MacGregor]

the mail changes look fine to me Benjamin.

Comment 3

[Benjamin Smedberg]

Not going to make 1.8 without a miracle. Shaver, do you need a miracle to do
lightning?

Comment 4

[Mike Shaver (:shaver emeritus)]

No, I have largely abandoned deeply-integrated search in Lightning 1.1, along
with participation in the threading model, because of this and other related
issues.  Makes me kinda sad, but I try to live in reality at least one day a week.

Comment 5

[Axel Hecht]

Comment on attachment 177745 [details] [diff] [review]
Add ownership models

Sorry, this has been taking way too long. Before I can actually review the
code, I need to understand the comment in the interface. And the dual ownership
is still confusing me. I opened http://wiki.mozilla.org/RDF:Delegates to get
closer to something without going thru attachements all the time. (We could
even morph that into docs on devmo, then.)

Comment 6

[Benjamin Smedberg]

Attachment: Whiteboard diagram describing the ownership models

Comment 7

[Mike Shaver (:shaver emeritus)]

Comment on attachment 177745 [details] [diff] [review]
Add ownership models

Delegating review to vlad, please stamp sr+ for me if you approve.

Comment 8

[[e-mail]]

Would it be worthwhile adding cycle collection to RDF resources?

Comment 9

[Benjamin Smedberg]

If we can require that all delegates and resources are cycle partipants, that would solve this problem nicely, yes.