Dave Cridland wrote:
> On Wed Oct  3 21:39:36 2007, Peter Saint-Andre wrote:
>> http://www.xmpp.org/extensions/tmp/xep-0048-1.1.html
> 
> Two comments I think need to be addressed relatively urgently:
> 
> 1) The document says it's defining a data format to store XMPP
> conference rooms and "HTTP URLs" - is there any problem with storing
> other scheme URLs? I can't see a reason for this restriction.

That was the original use case (in fact I don't know of any jabber
clients that bookmark web pages, but there may be some). I think that it
was not intended as a generic way to store any URI, but that it was
specific to web pages.

Naturally if we want a way to store any URI, we could write a separate
spec for that. :)

> 2) The format for storing XMPP conference rooms includes a password.

Password protected rooms are rare (typically people create members-only
rooms). I wonder if we even need this feature.

> This leads to two things:
> 
> a) The password may be exposed, if TLS is not used, etc, to a third
> party sniffing the connection. Although MUC uses the password in
> plaintext anyway, it seems likely to me that the password is likely to
> be more visible by this method.

Correct. That's bad.

> b) The password may be exposed to the server administrator - if it's a
> foreign administrative domain holding the conference room. Again, this
> exposure can happen anyway, if the MUC room is connected to, it just
> seems to me to making the problem worse.

Agreed.

> These should be mentioned in the Security Considerations, and we should
> consider alternative options, which may be:
> 
> i) Don't ever put the password element in - clients should handle the
> error on joining and prompt the user for the password.
> 
> ii) Do put the password element in, but leave it empty, and say that a
> zero length string as a password is a special case meaning that the
> conference room requires a password, and the client should prompt the
> user for it.
> 
> The latter mechanism might save a round-trip.

Might. But if I bookmark the room as requiring a password but the room
owner changes the room to no longer require a password, then we have
possible confusion.

So I would vote for (i) don't include the password elemenet (and
deprecated ASAP).

> More involved comments:
> 
> There's quite a wealth of prior art in the area of bookmark storage and
> roaming. Not only is there XBEL and existing browser formats, but
> there's also the ACAP dataset class for storing bookmarks. This is
> pretty readable even if you don't know about ACAP.
> 
> http://tools.ietf.org/html/draft-ietf-acap-book-06 - Section 4.2 is the
> one to read.
> 
> Both XBEL (which I only vaguely recall) and the ACAP dataset (which I've
> implemented) contain more than just title and URL. Some of this data has
> been overtaken by trends - a hierarchy of bookmarks is no longer
> ubiquitous - but a lot hasn't, such as a description, etc.

Thanks for the pointer!

Peter

-- 
Peter Saint-Andre
https://stpeter.im/

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to