[Standards] UPDATED: XEP-0237 (Roster Versioning)

Peter Saint-Andre stpeter at stpeter.im
Thu May 14 19:52:43 UTC 2009

Hash: SHA1

On 5/14/09 1:39 PM, Florian Zeitz wrote:
> Peter Saint-Andre schrieb:
>> Is this whole tempest in a tea pot about the *examples*?!? (Sorry, I
>> haven't read the entire thread.) If so, then I will change the values to
>> the following:
> Maybe it's not but that is what I take from it right now. Quite frankly
> I don't know why this thread has gotten so long, but here is a
> relatively short summary of what I now understand to be the "problem".
> (And also what is not the problem IMHO):

Summaries are always appreciated. Thanks!

> Not the Problem:
> Opaque versions: Curtis King seems to think restricting ver to numbers
> is a good thing for interoperability, but like you (stpeter) I don't see
> how. It's generally fine/good to make ver opaque.
> Suggesting hashes for 5.2: It is a possible implementation and feasible
> to do, even if possibly not the easiest.

I've already removed that "suggestion" and merely described possible
approaches in the implementation notes. I'll push out a new version of
the spec soon.

> The Problem:
> The flow in Section 3 shows roster pushes. Those require the server to
> know the order in which roster pushes happened (at least if I understand
> it correctly), which is probably the reason 5.3 specifies ver as
> strictly increasing numbers, so you can use ver as the "timeline".
> The Example shown here doesn't use numbers though, which is confusing
> (Jiri also seems to assume it may lead developers to write insane code.
> Maybe he is right, maybe not, but confusion should be avoided anyway).
> Example 3 from Section 2.3 is just a pain to read, because you have to
> compare long strings to understand the accompanying text. It also
> inhibits the same problem as Section 3.

Sure, that makes sense.

> The obvious solution would be to use numbers in this places. People have
> argued that this would also lead to bad client developer behaviour,
> because they might assume ver is always an integer. Matthew Wild
> suggested putting ver="<char><version>" in the examples. I personally
> think that's fine.

Works for me.

> I also hope you are not to annoyed by the fact that people seem to get
> hung up on the examples.

Nope, it's important to get the examples right.


- --
Peter Saint-Andre

Version: GnuPG v1.4.8 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org


More information about the Standards mailing list