[JDEV] Re: Server Independent Accounts (Michael Brown)

Michael Hearn mhearn at subdimension.com
Sat Jun 2 14:33:40 CDT 2001

Hi Michael,

Well, I don't know if it was considered in the early days of Jabber, but I
have to disagree with you for several reasons:

- Having a username at hostname address format is actually one of the few good
ways we know of to have a unique identity. That's why Microsoft Passports
are identified by email addresses: they are guaranteed to be unique, without
- It's a good way of avoiding centralisation, as any computer can be a
server without registering itself with any authority other than the DNS
- Yeah, I agree that being tied to one particular server is bad when you
want to move, but it's better than the alternative. Just imagine every email
user in the world competing in a single namespace. The Hotmail problem of
being called mikehearn30503 would be multiplied to the nth degree!

However, I agree with you that we need better ways of allowing people to
move between servers. So, here are some ideas:

- A redirect protocol which allows a server to automatically redirect
packets destined for one account to another transparently, possibly with
some kind of return warning that the account name has changed and that a
redirect has taken place. This should preferably be independant of the
server (ie can be set by a client) to stop an ISP holding you to ransom with
your Jabber identity.
- A way of downloading to a local file, then reuploading your roster
(INCLUDING private namespaces), so you can transfer all information that is
owned by you between servers.

thanks -mike

> Message: 6
> From: "Michael Brown" <michael at aurora.gen.nz>
> To: <jdev at jabber.org>
> Date: Sat, 2 Jun 2001 12:49:11 +1000
> Subject: [JDEV] Server Independent Accounts
> Reply-To: jdev at jabber.org
> I was trying to convince a friend to change to Jabber a while back, and he
> was telling me that Jabber wasn't really what he was looking for.  He is
> trying to find an IM service that has accounts that are independent of the
> server they are running on.
> How it would work is this:
> You are assigned a unique account name.  (Ok, coming up with unique names
> can be tricky, but bear with me).  For arguments sake, lets say mine was
> "Michael_S_Brown".  (Note there is no server as part of the address)  Now,
> when someone sends me an instant message, something similar to an DNS
> happens, and my account is resolved to a particular server.  Say I am at
> work at that particular time, it gets resolved to
> Michael_S_Brown at mywork.com - however if I was connected from home, it may
> resolve to Michael_S_Brown at myisp.com .
> One of the advantages of this, is that if I change jobs, people don't need
> to know that, and can still send messages to Michael_S_Brown.  They don't
> care if it resolves to Michael_S_Brown at mywork.com or
> Michael_S_Brown at mynewwork.com .  Likewise, if I change ISP's and I am at
> home, it might resolved to Michael_S_Brown at mynewisp.com but people
> me neither know nor care that I have changed ISP's recently.
> Having email addressed to a particular server was always a pretty lame
> although I'm sure the ISP's love it, because it makes it harder for users
> change to another service provider.  Think how hard it would be if your
> website was tide to a particular server, and if you physically moved your
> site to a different location you had to email everyone who might go to
> site and tell them "Sorry, but I have changed servers.  If you want to go
> http://www.mydomain.com now, you will have to remember to type
> http://www.mynewdomain.com".  It's not very practical, but it is exactly
> what we do with email addresses, and now Jabber addresses.
> Anyway, my question I guess is:
> Was anything like this considered in the early days of Jabber design?
> If so, was it discounted for some reason?  Why?
> If not, is there anyway to incorporate something like this into the
> Jabber spec?
> (I suspect not - the best thing I can come up with is some sort of message
> forwarding like bigfoot.com does for email).
> Michael.

More information about the JDev mailing list